November 30 marks the four-year anniversary of ChatGPT’s launch. In that time as a writing professor, I’ve seen much of the work I do change dramatically, and not for the better. The undergraduate courses I teach are riddled with AI writing, and I’m not alone.

The New York Times reported this summer that between two-thirds and 90 percent of high school and college students are using AI for schoolwork. A recently published study of 77 students by researchers at Virginia Tech found that only six students refrained from using AI to help them write, and nearly half of the students used AI and asked it to write their entire essays.

Individual teachers and institutions have come up with creative ways to overcome AI challenges by designing assessments that are inherently difficult for AI to ace, but collectively, the field of education, and higher education in particular, has not done enough to prevent AI misuse when it comes to writing. Instead, by and large, education has treated the problem of AI writing as an abstract and unconquerable evil, such as human suffering.

But the challenge of AI-generated writing in the classroom has always seemed like a fixable issue to me, if only there was a greater, widespread commitment to fighting it. Regardless of what you think of AI use overall, the idea that instructors should be able to ensure their students are actually writing what they say they are writing shouldn’t be controversial. Nor should we give up on trying to suppress the flood of AI writing.

Here are some steps institutions could take unilaterally or collectively to push back the ever-rising tide of AI.

1. Embrace AI Watermarks

AI watermarks act as a kind of digital fingerprint on AI work. The technology already exists to implement these, and it's likely soon be required for most major AI platforms, particularly to comply with the European Union Artificial Intelligence Act, which requires watermarking.

Educators and education leaders should loudly applaud these efforts and be prepared for when watermarks are rolled out in the U.S. Consequently, schools and universities should issue clear guidance for students and teachers on how AI watermarking will be tested by educators as well as what the penalties and procedures will be once an AI watermark is detected in a student's work.

2. Use Effective AI Detection Tools

AI detection tools emerged early as an obvious hedge against inappropriate AI use for writing. The problem is that many of these were initially terrible and resulted in false positives and nightmare accusations for students, which caused many in the field to immediately dismiss them. That was a mistake.

With the right procedures in place, AI detection tools can be utilized fairly. For example, Pangram has very high accuracy rates according to independent studies, and the company says its rate of falsely identifying a human-generated document as AI-generated is roughly 1 in 24,000. These numbers should continue to be assessed. If those rates hold up, there’s no reason this or other effective AI detection tools can't be used by educators.

3. Put Collective Pressure on Tech Companies

AI technology is evolving and no response to it in the school setting is perfect . . . for the corporations creating the platforms! Not so much for students and educators.

So far, the vast majority of AI misuse in school involves only a few platforms. Still, these solution providers need to take more responsibility in addressing what they’ve done to traditional writing assignments in schools.

If a new smartphone was released that could hack into school data centers and change student grades, educators would be outraged and demand these tools be altered to protect student learning. And yet, this is essentially what is happening with AI writing.

As educators, we should demand tech companies lead the way in ensuring their products don’t obliterate academic integrity. I’m not naive enough to think these types of efforts would actually be effective, but there’s something to collectively and symbolically letting the tech world know that they are causing problems and that we can fix these problems with their help.

4. Greater Emphasis on Originality in Writing

If there’s one bright side to the surge in AI-generated generic writing, it’s that it has demonstrated just how vital creativity in writing is. We should have been fostering this more long before the AI revolution, but it’s become increasingly more critical now.

Writing assignments should not be used merely as ways for students to demonstrate their learning, but as opportunities for students to show their creativity, as well as take conversations and ideas further and in new directions than previously. We should start weighing originality and unique research and perspectives more strongly in our rubrics.

Additionally, we should be unabashed about giving lower marks for work that either was written by AI or sounds as though it could've been written by AI. This might be somewhat subjective, but no more so than many aspects of writing we grade, such as tone or strength of argument.

By centering creativity and originality in our approach to writing and grading, we’ll be establishing a foundation for human writing that can hopefully help stand up to the current generation of AI tools and subsequent iterations. I don’t think we’ll ever return entirely to the pre-ChatGPT days of student writing, but with some of the strategies discussed here, we can get a whole lot closer.

Google Search has fully integrated AI through a feature called AI Overviews. For some time now, if you phrase your query as a question such as, “When is the next train from Brewster to Grand Central?” the tool will tell you the times the train is leaving . . . except I wouldn't trust that information.

According to one recent analysis conducted for The New York Times, Google AI Summaries are accurate about 9 out of 10 times, which tracks with my experience. Whenever I ask a series of questions around a subject I'm well-versed in, I almost always find minor errors and often notice outright falsehoods, not to mention the aforementioned inaccuracies with train schedule-related questions in my area.

This led me to realize that Google’s AI summaries can be a great tool for teaching AI literacy and giving students tangible insights into AI hallucinations, limitations, sourcing, and more.

To this end, I’ve designed a short assignment. I ask students to use Google’s AI Summary tool for a subject they know well. Then they have to ask it to provide a bio of them or information on their school or the time of their local train or when their favorite sports team is playing. I require that students keep querying until they find falsehoods, which can sometimes take a little time. Then I ask them to verify the sources and explain why and how AI got it wrong.

Though I’ve only recently implemented it, I find this AI exercise enforces the lessons discussed below.

1. Google AI Overviews (And AI Outputs In General) Need Verification

This is an obvious lesson, and one that students already know in theory, but in practice, I find they don’t apply it. Students, and many people in general, will insist something is correct because AI told them, even though they know AI gets things wrong frequently.

Forcing students to confront AI accuracy with this assignment can cut through some of the cognitive dissonance. It’s harder to dismiss AI mistakes when those mistakes are about you or something you know very well, such as your town or school.

2. Source Verification

When I asked Google what nonfiction author David Grann’s next book is, it said he is writing a new book about the Nobel Oil empire. When I clicked on the source to verify this, I was led to Grann’s social media page. As I explored, I realized he was not working on these books but had posted about books on these topics from other authors.

Having students go through a similar process with source verification provided by Google’s AI summary can teach them not only to better spot AI falsehoods but also about source verification more broadly. Even with human-generated work, sometimes when you go back to a cited source, you find it didn’t frame the claim the same way or, in some cases, that the author is making claims the source material can’t support.

Therefore, this type of fact-checking is a valuable skill for AI verification and beyond.

3. AI's Inner Workings

The mistakes this type of tool makes can hint at larger problems and concerns with AI. Any student who knows how to use the internet can look up train times and report back on arrivals and departures.

Powerful AI tools know how to search the internet with an efficiency beyond what any human could dream of achieving, yet these also can’t be trusted with the same simple task. This happens because AI engines don’t verify facts the way a human might. Instead, most current-generation models predict phrases and statements through pattern recognition. This has led some to refer to them as “stochastic parrots,” though that terminology is sometimes derided, correctly, I believe, as an oversimplification.

Either way, the fact that many AI models can’t be trusted to perform a relatively simple task is a good starting point for discussions about the strengths and limitations of these tools, and a direct example of how something about how these operate very differently than humans.

4. Learning When And How To Use AI

Despite all this, I’m not advocating for students to avoid using Google’s AI Overview or some other AI tool. When Google's AI Overview works, it can be a helpful. In fact, one of the main AI uses I’ve found personally useful is as an alterative to a search engine.

Google AI Overviews often get me the information I’m looking for quicker than a traditional search might, and I like seeing what it comes up with before I embark on a traditional search. I’ve just learned not to trust it and always to verify what it says, even on small, seemingly trivial items, such as when the next train is arriving at the station.

Finding these uses and learning the limitations of AI is something important for students to realize as well. Whether we like it or not, we're all interacting with AI content frequently, and becoming more skilled at critically absorbing that content is a necessity for everyone.

Engaging in this type of conversation with students has helped me better focus my own AI use, and I hope has done the same for them. Now, I just need someone to help me figure out my local train schedule.

AI watermarking could soon provide a way for educators to tell whether students have used AI or not.

To comply with the European Union Artificial Intelligence Act, tech companies will have to watermark content generated through their AI tools. Google is already watermarking text created through its products, and Anthropic will start doing so with similar techniques. OpenAI is expected to develop watermarking methods for its products as well.

Sarah Fisher, a lecturer in philosophy at Cardiff University in Wales, believes this is all a step in the right direction inside and beyond the classroom, though she acknowledges the new technology won’t be perfect.

“As a society of humans, we place value on what humans produce,” she says.

What Is AI Watermarking?

Historically, watermarking has involved some type of visible marker of authenticity.

“You watermark your banknotes, and you know that they're not counterfeit, or hopefully not,” Fisher says.

AI watermarking works somewhat differently, though the end goal is the same. Claude’s watermarking system will make the chatbot follow specific statistical rules for choosing words with the same meaning. Human readers won’t notice anything, but someone with the watermark key will be able to see if the string of words Claude uses is consistent with the watermark, Fisher explains.

“So, if I'm a student, and I go to my ChatGPT, or my Claude or whatever, and I say, 'Write me a hundred words about X, Y, or Zed.' I wouldn't necessarily see a watermark in what gets produced, but if I then submitted that to an online submission system, then presumably, at some stage, we'll have some software that would be able to tell that it was generated by ChatGPT or Claude or whatever.”

Why Could AI Watermarks Be Good News For Teachers?

Fisher believes being better able to tell what is generated by AI vs. humans will be helpful both for educators and society at large, for a few different reasons.

“For some categories of content, it's really obvious,” she says. “If we're talking about especially audiovisual stuff, these deepfakes are so insidious, we really want to know whether what we're seeing before our eyes is something that really did happen.”

Identifying AI text is also important, but in different ways.

“When we get a piece of text, it's not like looking at a deepfake," Fisher says. "We already have to assess whether what we're being told is true or false. And yet, I think it still can be really, really important to know whether that bit of text came from a human or a machine.”

Fisher adds, “One reason is informational or epistemic. These Chatbots tend to make some signature errors that humans don't make and vice versa. So it's just useful to know what to look out for.”

Another factor is just the value of protecting human work. “I've heard a lot of people saying things such as they just don't want to read a story that's written by an AI," says Fisher. "Even if it was the same words as a human produced, something's missing.”

Discerning human-generated work can be particularly important in the classroom setting.

“As educators, we want to know what our students are producing," Fisher says. "We don’t really care about the end products that much. We care about them going through the process. It's not like we're filling a demand for essays in the world.”

Will AI Watermarks Slow The Tide of AI In The Classroom and Beyond?

Fisher believes the answer to this question is yes, but only to a point.

“I think it might help a bit. I don't think it's going to be the silver bullet that solves this problem, and maybe there just never is a solution to this problem,” she says. “There's always a small minority of students who will find a way to cheat. And that's unfortunate, but we should never expect to get it down to zero. But I think the watermarking might help with some of the most egregious cases, where if a student literally has generated a whole long-form piece of text and taken it straight from Claude, say, it seems like in the future the watermark would be pretty good at telling the educator, ‘Look, this is extremely likely to have been generated by Claude,’ and it's very unlikely that we'd get false positives through that.”

What Are Some Limitations To AI Watermarking

It’s already certain AI watermarks won’t catch all AI-generated content.

“There might be rogue systems which aren't doing the watermarking,” Fisher says. “We know that the big companies are trying to do it, but these tools are proliferating, and so if a student has used one that doesn't have a watermarking feature, then that may not get caught.”

Additionally, existing watermarking techniques are less good at catching short passages of AI-generated text.

“Companies are being quite open about the fact that it might be possible to edit text in a way that takes out the watermarks,” Fisher says.

Though she adds, “I don't know how worried to be about that, actually, because once a student's gone to that much trouble, maybe they've done the work by then.”

Future of AI Watermarking

Some critics of AI watermarking worry that since it won’t be perfect, it will lead society to a false sense of security and could potentially make it more likely to fall for AI-generated content, Fisher says. Another criticism is that past efforts to label traditional misinformation on social media have had mixed results, and some think the same thing will be true about labeling AI.

Fisher doesn’t dismiss these concerns, but believes on the whole, AI watermarks will be a good thing.

“I'm reasonably optimistic that this is a positive step in the right direction,” she says. “It's helpful for us all to hold humans and AI apart for now while we figure out exactly what these things are and how we're going to integrate them into all of our social practices and our institutions.”

Ray-Ban Meta glasses are a type of wearable tech that can take photos, record, play music or podcasts, capture audio, and pair with AI so you can do a lot of things hands-free. The more expensive Ray-Ban Meta Display glasses also feature an on-lens display meant for tasks such as giving you text updates without blocking your view of the real world.

I recently tried the basic model (without the display) and found that these offer some compelling lesson planning and video creator capabilities for teachers. These also have potential for students and could aid in translation, accessibility, and more, but the price of the tech--the third generation starts around $450 per pair--is currently prohibitive in many education environments. Significant privacy issues are also a concern with the devices.

Here’s a closer look at the Ray-Ban Meta glasses and what educators should know.

What Are Ray-Ban Meta Glasses?

Ray-Ban Meta glasses have also been referred to as AI glasses or smart glasses. These offer a chance to put your phone and much of what it can do in POV mode, allowing you to capture high-resolution photos and videos that show what you see. These also function as a sleek headset with good calling and listening features. More advanced models also offer displays. Different styles of sunglass are available, such as aviators.

The Ray-Ban Meta glasses are a rare piece of VR or augmented reality AR tech that seems to be a success. Ray-Ban’s parent company said that more than 7 million pairs of the VR glasses were sold last year, more than tripling sales numbers from the year before.

What Are Some Ways Educators Can Use Ray-Ban Meta Glasses?

The main use I see for Ray-Ban Meta glasses right now are as an alternative to a wearable camera such as a GoPro. If you’re an educator who films a lot of lessons for students, maybe as part of a flipped classroom setup, these glasses could be a worthy investment. It’s really cool to just film what you’re seeing, and I could imagine using this for lab prep, or to illustrate a lesson. You could also give your students a more immersive quick video tour of a historic site.

Influencers have embraced these glasses for activities such as cooking demos, but for those of us who are not would-be celebrities, these might be better used to provide a helpful way to film practical demonstrations.

Accessibility Potential

In addition to making videos, Ray-Ban Meta glasses seem to offer a lot of potential for accessibility in education. A translation feature worked really well in my tests of Spanish-to-English, but it still provided a text-based translation. The person you’re speaking to’s voice is captured by the glasses, but the translation of what they say appears on your phone. This was convenient but not that big of an improvement over a phone-based translation app.

That said, it’s easy to imagine future generations of this technology adding more robust translation features: it would be nice to have the translated text appear on the glasses themselves, for instance.

Even right now, students could, in theory, use the glasses to photograph slides and info written on the whiteboard or chalkboard. For those with vision impairments, the screen display version of the glasses could allow them to see text up close. However, the technology is still too much in its infancy for these uses to feel really practical today.

The cost is also prohibitive, and most importantly, my concerns about the privacy features of these glasses prevent me from recommending these for student deployment.

Privacy Concerns

The privacy issues with Meta Ray-Ban glasses are significant both inside and outside of the classroom. The New York Times reports that restaurants have banned the glasses, and there are reported incidents of these being used to harass people at gyms.

Some users figured out how to disable the red light that turns on when the glasses are recording; in response, Meta announced that it had permanently disabled the recording abilities of thousands of device owners who had tampered with the glasses in this way. This has led the devices to earn the unflattering nickname of "pervert glasses."

None of this has stopped Meta from moving ahead with the device, though they now offer a record-only version of the glasses. Nor is it preventing other companies from releasing similar devices. Google has a new smart glasses tool in the works, and there are rumors Apple does as well.

In a school setting, however, all these privacy issues would be compounded. On top of inappropriate recordings, there are legitimate concerns about how the technology works with Meta’s AI tech and how student data and safety is addressed. Those are always worries with student tech, but a wearable device that sees and hears everything a student does makes these already vital concerns even more pressing.

Bottom Line: Ray-Ban Meta Glasses And Education

Right now Ray-Ban Meta Glasses are a neat, if somewhat gimmicky, device that would be good for educators looking to make POV educational videos.

Beyond that, these have a lot of potential but are not ready for primetime just yet from a tech capability standpoint. When you add in the significant and quite serious privacy concerns, these aren’t devices I see being handed out in a school anytime soon.

That said, the questions around privacy and smart glasses are ones society will need to deal with in the coming years and, as such, it's worth familiarizing yourself with the devices and discussing these with your students, even if Ray-Ban Meta glasses are something you’d personally never wear.

Every week, the headlines seem to bring a new iteration of alarm regarding generative artificial intelligence in education. From fears of widespread cheating and the death of student writing to warnings about cognitive atrophy, algorithmic surveillance, and the erosion of intellectual independence, the public dialogue is rife with anxiety.

When headlines sound like an impending apocalypse, educational leaders are left standing at the crossroads of community trepidation and educational reality.

Yet, for educational leaders navigating this disruption, the first and most critical leadership act is not reaching for digital blocking software. Rather, leaders should make sure they obtain a historical perspective and consider the history of concerns over technological change. By contextualizing the modern media panic within the history of drama and panic over technological change, leaders can help to guide their institutions though the rough and fluid waters as generative AI grows to maturity within the educational arena.

The Familiar Cycle of Innovation Panic

Every time a paradigm-shifting technology enters human culture, widespread resistance follows. Media narratives tend to frame emerging tools as existential threats to the intellect, but history tells a strikingly repetitive story.

Around 370 BCE, as recorded in Plato’s Phaedrus, Socrates lamented the invention of writing itself, warning that it would “create forgetfulness in the minds of those who learn to use it, because they will not exercise their memory; they will trust to the external written characters and not remember of themselves.”

While traditionalists worried that inscribed speech would hollow out human cognition, advocates such as Isocrates and Athenian democratic institutions recognized writing as an indispensable tool for composition, transparency, and civic progress. The modern concerns with cognitive atrophy seem oddly attuned to Socrates’s arguments against the alphabet.

Centuries later, similar patterns unfolded:

• In 1913, Thomas Edison proclaimed that printed books would be obsolete within ten years, completely replaced by motion pictures.
• By the 1920s and 1930s, cultural commentary warned that the radio would disturb adolescents’ excitable minds and turn education into “a matter of pressing the button.”
• In the mid-1970s, mathematics educators sounded the alarm that pocket calculators would erode basic arithmetic and ensure children would be helpless whenever batteries went dead.
• In the 1980s, historians and critics warned that microcomputers were vastly oversold and posed an unfounded distraction to traditional pedagogy.

Outside classroom walls, leaders routinely dismissed technologies that reshaped the globe, from Western Union calling the telephone an inferior toy in 1876, to hardware executives dismissing personal computing and the internet in the late 20th century.

When educational leaders view today's headlines about generative AI through this lens, panic softens in perspective. AI is not an unprecedented end to human intellect; it is the latest chapter in our long history of renegotiating the boundaries between cognitive labor and cognitive tools.

Moving From Reactive Bans to Guided Literacy

When generative tools such as large language models emerged, an instinctive reaction across many school districts was simple: block and ban. However, as panels at leadership forums such as during Tech & Learning’s Texas Regional Leadership Summit have made clear, a restrictive approach solves nothing.

Generally, blanket restrictions simply widen the digital divide and force students to experiment in unguided, shadow environments. Instead of isolating classrooms from emerging technologies, educational leaders must step forward to guide students toward ethical, real-world innovation. As Ethan Mollick notes in Co-Intelligence, homework cannot be preserved by pretending large language models do not exist; students must learn to treat AI as a tireless collaborator and thought partner while educators raise academic performance baselines.

Reframing Assessment: The Learning Process Over the Final Product

To address concerns about academic integrity and student agency, schools must fundamentally rethink what they measure as assessment of learning. Educational leaders must guide instruction away from task-completion models and the practice of grading outputs. The entire learning process will need to become more reflective and interactive. Students will need to be asked to consider how they are learning and how successful the use of generative AI tools are within the scope of each assignment.

When assignments measure only a polished five-paragraph essay or a completed worksheet, an AI model will always offer an enticing shortcut. Educators need to prioritize the learning and design process, feedback loops, and deeper conceptual understanding, and AI becomes a catalyst rather than a cheat code. As Salman Khan argues in Brave New Words, generative models can serve as personalized, one-on-one tutors and instructional assistants, addressing Bloom's classic two-sigma problem by supporting deliberate practice.

Simultaneously, district policies must heed the cautious recommendations of the U.S. Department of Education and international bodies such as UNESCO. Leaders must maintain a strict "human-in-the-loop" approach, ensuring that algorithmic models do not usurp cognitive agency or institutionalize evaluation biases. All processes need to enforce the central need for human agency and judgement within the instructional or creative process.

The Path Forward for Leadership

Headlines will continue to oscillate between techno-dystopian terror and commercial hype. The task for education leaders will be to remain grounded.

By understanding the historic rhythm of educational technology panics, leaders will be able to more effectively guide their institutions through this time of rapid change.

The Point Where I Used to Stop

For most of my life, I assumed coding belonged to other people. It was for developers, programmers, and those who understood what all those brackets and symbols meant. I had taken AP Computer Science in high school, but that experience only reinforced the idea that coding was not for me. I remember making it through “Hello, World!” before becoming lost somewhere between functions, variables, and whatever came next. Whether “Hello, World!” was an assignment, a function, or simply a warning sign, I am still not entirely sure.

After that, I stayed away from coding. I could appreciate what technology made possible, and I had dabbled with formulas and macros, but I still did not see myself as someone who could build anything with code.

What I have never been good at accepting, though, is an inefficient process. When I run into a repetitive administrative task that feels more complicated than it should be, I usually start looking for a better way to do it. I search the web, try different features, look for extensions or add-ons, and, more recently, ask AI for ideas.

Quite often, I eventually reach the same answer: the solution involves code. That is usually where I stop. I still cannot sit down with a blank screen and independently create a Google Apps Script, Google’s tool for using code to automate tasks across apps such as Sheets, Docs, Drive, and Gmail.

What I can do is identify a frustrating process, explain what I want to happen, follow clear directions, and keep asking questions when something does not work. As I learned during this experiment, that combination can be enough to get surprisingly far.

One day, I wondered why I was stopping at that point. If AI could help me find the solution, could it also help me write the code? That question became an experiment.

The Annoying Problem That Got Me to Try

My experiment grew out of a task that was not especially difficult. It was simply annoying.

Whenever I needed to share materials with a principal, colleague, or student, I typically would create a folder, name it, grant access, copy the link, send it, and then keep track of where everything was located. Doing that once was manageable. Doing it over and over was the kind of small administrative task that quietly eats away at your time.

Normally, I would have searched for a tool that could help. This time, I described the entire problem to Gemini instead. I explained that I wanted to start with a spreadsheet containing names and email addresses, create an individual folder for each person, assign the appropriate access, and have the folder link automatically written back into the spreadsheet for my reference.

Gemini generated the code, but the code by itself was not what made the experiment possible for me. It also showed me where to open Apps Script, where to paste the code, what I needed to change, and what to click.

When an error appeared, I copied the message into Gemini and asked what it meant. When the script did something I did not expect, I described what happened and asked it to fix the problem. I did not suddenly understand programming; I simply kept following the directions and asking the next question.

Eventually, the tool worked. I could identify a main folder, add names and email addresses to a spreadsheet, and run the script. It created an individual folder for each person, assigned access, and placed the link back into the spreadsheet. The same basic idea could also be adapted to create individual Google Docs, Sheets, or Slides from a template.

Of course, there are other ways to solve a problem such as this. Extensions, add-ons, and tools in the Google Workspace Marketplace may already do some or all of these things, and sometimes may even be the better solution.

What was different this time was that I was not trying to fit my workflow into someone else’s tool. I started with the workflow I wanted and built the tool around it. I could decide what information appeared in the spreadsheet, how folders were named, what the script did, and what information it returned to me.

For a developer, this may be an ordinary project. For me, it represented a significant shift. I had moved from thinking, “There must be a tool for this somewhere,” to realizing, “Maybe I can build the tool I actually want.”

Before using it with real recipients, however, I tested it with sample accounts, checked the folder locations, and verified the permissions. A mistake in an email address or sharing setting could give access to the wrong person, so the ability to automate something did not remove the need to check my work. I also asked a friend who teaches computer science to review both the tool and the code to make sure I was not missing anything important, and I made sure everything aligned with my organization’s policies before considering it ready to use.

Once I Built One Tool, I Started Seeing Other Possibilities

Getting the first script to work changed the way I looked at other repetitive tasks. Instead of immediately accepting a cumbersome process, I wondered whether I could build something to make it easier.

One example involved receiving a large master spreadsheet containing information that needed to be separated and distributed to different people. The manual process required filtering the file, copying the correct rows, creating separate documents, reviewing each one, and preparing the corresponding emails.

I asked Gemini whether a script could handle some of those steps. Together, we developed a tool that could separate the spreadsheet into recipient-specific files and prepare the corresponding emails. Rather than spending my time repeatedly filtering, copying, and creating files, I could spend more of it reviewing the results.

This example also showed me where the limits of my experimentation needed to be. An incorrect filter could place the wrong information in a file or send it to the wrong recipient. I would avoid using an AI-generated tool such as this with student-level data because the consequences of an error could be significant. I would limit it to information that could be safely distributed and that complied with my organization’s policies.

I also preferred to have the script create email drafts rather than automatically send anything. That gave me the opportunity to check the recipient, attachment, subject line, and message before anything left my account. Saving time was useful, but not at the expense of judgment.

Another experiment came from a similar frustration. I often work from spreadsheets containing names, roles, schools, and email addresses, so I wondered whether the spreadsheet itself could become a more useful communication tool.

With Gemini’s help, I created a script that allowed me to select people from the spreadsheet and decide how I wanted to contact them. Depending on the situation, I could place recipients in the To, CC, or BCC field or create an individual draft for each person. The subject line and message could also be pulled directly from the spreadsheet.

None of this is revolutionary technology, and that is actually part of the point. I was not trying to create the next great software platform. I was solving small problems in my own work and shaping each solution around how I actually worked.

The Skill Was Not Really Coding

Someone who knows how to code would already understand functions, debugging, onOpen(), and many of the other concepts that still feel unfamiliar to me, and they would almost certainly produce cleaner and more sophisticated code than I could. But that was not the expertise I brought to the experiment.

What I brought was a clear understanding of the problem, a solid grasp of the workflow, and a sense of what I wanted to happen. I could recognize when the result was wrong, describe the desired outcome, follow instructions, test what AI produced, and explain what needed to change.

Those abilities are not unique to programmers. Educators and school leaders use these constantly. We identify problems, explain processes, coordinate people, review information, and revise plans when something does not work.

Working with AI made me realize that those skills could also be part of building a technical solution and could help democratize solutions to everyday problems. I did not need to understand every line of code before I could experiment with it, but I did need enough understanding to test the tool, question the results, recognize the risks, and know when I was moving beyond what I should attempt on my own.

AI Did Not Make the Problems Disappear

Of course, the experience was not as simple as telling Gemini what I wanted and immediately receiving a perfect tool.

Sometimes a column heading was wrong. Sometimes the script could not find the folder. Sometimes it created duplicate files or stopped before finishing. At other times, I received an error message that meant almost nothing to me.

Previously, that would probably have been the end of the experiment. This time, I went back to Gemini, described what happened, shared the error message, and asked what needed to change. Sometimes the repair worked immediately, while other times it introduced a different problem that required another round of questions.

I was still debugging. I just had help doing it.

AI did not remove the frustration of getting something to work, but it changed what happened when I became stuck. Instead of reaching the limit of what I knew and abandoning the idea, I had somewhere to take the next question.

The Guardrails Still Matter

None of this means that someone should copy AI-generated code, approve every permission request, and hope everything works. Schools and districts may have policies governing Google Apps Script, external AI tools, automated emails, data access, and file sharing, and those policies need to come first.

Privacy matters during development as well. Gemini does not need actual student names, email addresses, identification numbers, grades, or confidential district information to help me write a script. I can describe the structure of a spreadsheet, provide column headings, and use invented names and sample data.

Testing matters just as much. Before using a script with real recipients, I can test it with sample accounts, review permissions, inspect drafts, and confirm that the correct files were created. When the stakes are higher, having someone with more technical expertise review the code can provide another important layer of assurance.

AI can help me debug the code, but it cannot take responsibility for what I choose to do with it. That responsibility remains mine.

Moving Past the Place Where I Used to Stop

What surprised me most during this one-day experiment was not that AI could write code. I already knew that. What surprised me was what happened when I stopped treating code as an automatic dead end.

For years, my question was simple: “Do I know how to code?” Since the answer was no, that usually ended the conversation. AI has given me a different question: “Can I clearly explain the problem I am trying to solve?”

That shift matters. AI does not make me a developer, nor does it replace the expertise of people who actually know how to code. What it does is give me a way to experiment with everyday problems that I once assumed were beyond my technical reach.

I am still not sure I have fully recovered from AP Computer Science. But now, when the answer to a problem involves code, I do not automatically stop there. I might just ask the next question.

A group of speech pathology graduate students are standing in front of a “patient." They listen as she tells them about the difficulties she’s encountered since returning to work after getting treatment for cancer.

“Everyone tells me I look great,” the patient says, “but I still don't feel like myself.” She adds that she gets tired easily and has started to cough when she eats, which she finds embarrassing.

One student asks her how often she gets the coughing sensation.

“Usually if I drink too fast or I’m not paying attention,” she says while looking at the student. She also provides more details about her symptoms.

The “patient” is an interactive AI hologram speaking from within a lifelike holographic display, a device about the size of a refrigerator that creates the illusion of a 3D and life-size person. It’s all part of a pilot program at the University of Central Florida designed to utilize holographic AI avatars to help give students the experience of interacting with actual patients.

Bari Hoffman, Associate Dean, Clinical Affairs and Healthcare Innovation, is leading this effort and is a proponent of hologram use in education. In 2024, she spoke with Tech & Learning about how holograms can provide lifelike lessons in STEM and other fields. The tools allow educators to “beam” in experts to speak in class, display intricate machinery, and provide more immersive examples of various kinds.

Hoffman has also recently begun using the technology in coordination with AI, a pairing that she believes will have significant impact for her students.

AI Holograms

The term “AI hologram” conjures all kinds of science fiction imagery, but in practice it can be a way for students to get more real-world, hands-on experience and be better prepared for careers. Students, such as the ones in the YouTube clip above, get a chance to learn more about what it’s like to interact with a real patient.

Traditionally, health students at the UCF have had the opportunity to interact with real patients who volunteer to come to campus, but such an opportunity is limited out of necessity.

“Somebody who has maybe some mobility challenges or other health-related challenges, it isn’t necessarily safe or easy for them to come to campus,” Hoffman says. So AI and hologram technology have allowed the school "to create something that is reproducible across programs, across days, across semesters; it allows us to really scale those experiences,” she says.

This principle extends beyond just working with AI avatars. Hoffman’s institution has also created realistic 3D videos of actual patients discussing their treatments and struggles. These stories can be replayed via hologram technology, and AI can make these even more accessible through its ability to translate patient testimony (with permission of course) into various languages that can reach students from different backgrounds, and be used for other educational purposes.

“We’ve been able to take our existing content and really translate that in any language, and that's allowed us to scale application and integration in a more culturally sensitive way, and even to expand our work in collaboration internationally,” Hoffman says.

The Power of The Hologram

Hologram technology goes beyond what students might glean from an AI avatar alone on a standard computer screen, Hoffman says. One of the reasons is what is called “copresence.” Holograms seem to fill up the room the way a real person might.

“You get that sense of realism, that eye contact," Hoffman says. "When you create your content properly, the content kind of has this Mona Lisa effect, where anywhere you are in the room, those eyes are going to follow you."

For example, Hoffman points to pediatric physical therapy students. “The students are able to do assessments with children two months of age to 10 years of age, and looking at individuals with low muscle tone or spasticity, that volumetric image allows the students to appreciate the center of mass and subtle changes in muscle tone that often are lost on traditional 2D platforms,” she says.

These types of realistic insights for students are part of why Hoffman's become such a strong proponent of the technology. She regularly hears from educators at other institutions and even some K-12 leaders looking to learn more about how hologram technology might fit into their institutions. She always advises teachers to find faculty champions, and make sure to keep learning objectives and the student experience in mind.

Hoffman also urges educators who are using holograms to think about how the technology can be used to create meaningful experiences for students. Then it’s important to measure that experience in some way, either through objective or qualitative data, that shows how holograms were used to achieve something new or different.

“That's got a lot of ROI to faculty, to programs, and to institutions,” she says.

Much of the conversation about students and AI has focused on cheating. Even esteemed tech reporter and columnist David Pogue recently framed student AI use through that lens in an online exchange we had. But what if we shifted the conversation from detecting AI use to thinking about the learning itself?

Instead of focusing on whether students used AI to complete an assignment, we can ask two more useful questions: Why are we giving students assignments AI can easily complete in the first place? And how might students use AI and/or other tools to support their thinking and learning?

Multiple-choice tests. End-of-chapter questions. Worksheets. Book reports. Generic essays. These assignments were problematic before generative AI. AI just magnified the problem and made it much harder to ignore.

That's why Michael Hernandez's work on Uncheatable Assessments caught my attention. Rather than focusing on catching students after the fact, Hernandez turns the lens back on us. What are we asking students to do, and is it actually worth doing?

Put Your Assignment to the Test

Hernandez created a tool that gives educators an Assignment Vulnerability Report. His Cheat Vulnerability Index scores assignments from 1 (vulnerable) to 10 (cheat-resistant).

The report looks at things such as originality, AI resistance, evidence requirements, personal or local connections, scaffolding, modality choice, authentic audience, reflection, and what we are actually assessing. Hernandez explains more about the thinking behind this approach in I Used AI to Build AI-Resistant Assignments.

I tried the tool with several assignments. A traditional worksheet earned just 1.3 out of 10. It lacked authentic purpose, evidence, meaningful choice, and personal connection. AI could complete it easily, but the real problem was the assignment.

Creative Doesn't Automatically Mean Uncheatable

A stronger assignment that gave students multimodal choices and asked them to explore human creativity and AI jumped to 5.9. It scored well for AI resistance, originality, and modality choice, but poorly for personal connections, authentic audience, and reflection.

That's important. Swapping a worksheet for a podcast, video, or presentation isn't enough. Students need something meaningful to do and a reason to do it.

This is something I've written about for years. In 5 Innovative School Models You Might Not Know, I looked at schools where students pursue interests, work on projects, learn in their communities, and demonstrate learning through real work rather than just tests and assignments created for the teacher.

What Might an Uncheatable Assignment Look Like?

Let's say students are studying pivotal moments in recent history. We could assign questions, a test, or an essay about how an event affected Americans. AI can do all of that pretty well.

Or we could ask students to interview an adult about pivotal historical moments they lived through. Now the student has something AI doesn't have: that conversation. They have to listen, ask follow-up questions, connect history to lived experience, and decide what matters.

Then give students agency over what they create: a podcast, picture book, video, biography, artwork, song, or another format. Give them examples and a rubric so they know what strong work looks like without requiring everyone to show learning the same way.

The audience matters, too. The person whose story the student is telling is an obvious place to start for finding an authentic audience. Students could also share their work with classmates, families, or the larger community. As I've written before in 5 Steps to Help Students Connect With Real Audiences, audience shouldn't be something we figure out after the work is finished. It should be part of why students are doing the work in the first place.

Reflection can be simple: What surprised you? What are you most proud of? What do you want others to understand? How did feedback affect what you created? Now we're seeing the thinking behind the product, not just grading what landed in the teacher's inbox.

I put this assignment into Hernandez's tool too. It scored 7.8 out of 10, including a 10 for local/personal anchoring, 9 for originality and authentic audience, and 8 for AI resistance and evidence. Because students gather primary source material themselves, the feedback called it "genuinely hard to fake or automate."

It still wasn't a 10. The tool flagged clearer expectations, stronger reflection, and more flexibility in how students present their work. I had already envisioned multiple formats, but making those choices explicit would make the assignment stronger.

Uncheatable Doesn't Mean AI-Free

Here's another thing I like about Hernandez's approach: uncheatable doesn't mean AI-free. He used AI to help build a tool for creating AI-resistant assignments.

Students could use AI to brainstorm questions, organize ideas, consider formats, or get feedback. AI can help students think more deeply without replacing their thinking. The goal isn't to prevent AI use. It's to create learning students can't outsource to AI.

AI happens to be very good at the recall, regurgitation, and generic production schools have assigned for generations. Those problems existed long before AI, which is part of the argument I've made about why we need to stop romanticizing school before technology.

AI didn't create the problem; it exposed it. We can spend our energy trying to catch students and preserve outdated assignments, or we can create work students actually have a reason to do. Then educators can stop trying to detect AI and start creating better assignments.

When schools make decisions about artificial intelligence, the conversation is usually led by adults. Administrators develop policies, educators establish classroom expectations, and technology leaders evaluate all of the platforms. These perspectives are essential, but I believe there is another important voice that is often included only after the decisions have been made, and that is the students who are already using AI.

Over the past year, I’ve had the opportunity to work with educators, leaders, students, and families across 12 school districts as they explored digital wellness, responsible technology use, and emerging questions about AI. Through student conversations, classroom visits, and task force planning sessions, I was reminded that students do not simply need to hear our guidance–their voice can help us create better guidance.

Students in third, fifth, and eighth grades, and in high school, offer insights that might otherwise be missed. They see AI used for homework, projects, online searches, and social interactions. They understand the pressure to finish work quickly and the temptation to seek help when an assignment feels confusing or overwhelming.

That does not mean students understand every risk. Knowing how to access an AI tool is not the same as knowing what happens to the information entered into it or why its responses may be inaccurate or biased. But listening to students can reveal where adult assumptions differ from their reality.

Several themes stood out in our conversations:

• Students want transparency. They want teachers and schools to disclose when and how AI is being used, especially when it involves grading assignments, evaluating their work, or generating feedback.
• Students want to understand how AI works. There is often an assumption that because students are comfortable with technology, they automatically understand AI. Knowing how to access a tool is not the same as understanding how it produces information, uses data, makes mistakes, or may reflect bias.
• Students want guidance. They want educators to model how AI can be used safely, ethically, and responsibly. They need opportunities to learn how to protect personal information, evaluate AI-generated content, and decide when AI should or shouldn’t be used.
• Students want clear expectations. Students want to know what is allowed for each assignment, which uses must be disclosed, what work must be completed independently, and where the boundary between assistance and cheating lies.
• Students want consistency and equitable opportunities. Confusion and frustration arise when one teacher permits AI use and another prohibits it without explaining why. Expectations do not have to be identical across every subject. Students should not have to guess what the rules are from class to class. If students in one biology class are permitted to use AI in meaningful ways, students in the same course with another teacher should have comparable learning opportunities, or it places them at a disadvantage.
• Students want meaningful learning experiences. They do not want to use AI simply because it is available or new. Several students shared with me that when they first started using Chromebooks, it was the best thing ever, but with repeated use without purpose, it made them want technology-free activities. AI should add value, and its use should be connected to a clear learning purpose.
• Students want protection from unfair accusations. Unclear rules can leave students worried that their own work will be questioned. Students need opportunities to show their thinking and demonstrate their learning.

Students can also help us understand the difference between the AI use we see and the AI use we do not. Educators might notice AI-generated language in an essay, but students are using these tools for more than schoolwork. Some use AI to seek advice, generate images, rehearse difficult conversations, or ask questions they are afraid or uncomfortable asking another person. Taking time to listen to students will help schools think beyond academic integrity and consider data privacy, well-being, relationships, and digital citizenship.

When thinking about creating policies, students can have a positive impact when they are involved in the conversation, be it through surveys, focus groups, technology advisory groups, or opportunities to give feedback on proposed guidance. Educators can also incorporate regular reflection into classroom activities and ask students how they used AI, what it helped them do, what was not possible with AI, and whether they had any concerns while using it.

Conversations such as these will provide educators with valuable information while also helping students to develop their AI literacy skills. Involving students shows that their perspectives are valued. Getting their input will help us understand how they are being affected by AI. When they have a comfortable space to share what they are seeing and experiencing and explain what they need, it leads to a stronger culture for learning about and working with AI.

If we want to prepare students to use AI thoughtfully, we need to ask what they are experiencing, listen to their answers, and let what we learn inform the decisions we make.

Common Sense Media’s new online, self-paced course examines AI use and its connection to student wellbeing.

While most discussions around AI in the classroom focus on topics such as AI tutors or AI cheating, an arguably more pressing concern is how these tools are impacting student wellbeing. More and more young people are turning to these tools for the type of mental health advice traditionally offered by a friend or therapist.

Common Sense Media’s free, hour-long class focuses on these types of AI use. I recently took the course and it was easy to follow, helpful, and a very productive use of an hour. The course lays out how students are increasingly turning to AI chatbots for emotional support, and explains why that can be harmful to their wellbeing. Then the AI and Student Well-Being course offers specific and actionable advice for educators to help their students overcome these challenges.

While some of the discussion in the course was familiar to me, other aspects of it were new. The whole course is a great reminder that I, and other educators, should be taking this issue seriously. Here’s what I learned.

1. AI and Student Well-Being: Students Are Using AI -- A Lot

In the course, Common Sense Media shares results from a survey it conducted in which more than a thousand 9- to 17-year-olds across the country were asked to discuss their AI use. It’s no surprise that the survey revealed that students are using AI a lot, but even so, the numbers are eye-opening.

“Nearly 9 in 10 kids age 9 to 17 use or interact with AI, including about one in four who do so on a daily basis,” the course reveals. Additionally, these regular AI users exhibit signs of AI dependence. “Twenty percent of kids who use AI, and 42% of kids who use AI daily, say it would be very or somewhat hard for them to go without AI for a month.”

Concerningly, students who said they were lonelier were also more likely to use AI more often and to turn to it for emotional support. Only 33% of kids who rarely use AI said they felt lonely some of the time. Meanwhile, “about half of kids who use or interact with AI daily (54%) or weekly (47%), and 44% of monthly AI users, say they feel lonely at least some of the time.”

2. AI Can Appear Helpful for Emotional Support

The course is not actually anti-AI and makes it clear that AI can appear (and might even be) helpful for emotional support in some instances. For certain students, AI could reduce feelings of loneliness in the moment, offer a low-pressure space to ask questions, and provide guidance and practice for navigating social situations. It could also spark creativity and self-expression.

But there are limits to these benefits. “[AI] does not actually know a student, care about their future, or share in the relationship,” the course text warns. “Its responses are generated, not felt.” This could lead to AI stopping students from seeking help from humans and its constantly affirming responses can reinforce “validation-seeking patterns.”

Additionally, “for vulnerable students, AI may intensify rather than ease existing struggles,” and, “One-sided interactions can start to feel safer or more satisfying than real relationships.”

3. Chatbots Appeal to Students Because They Offer “Frictionless Friendships”

One of the reasons AI chatbots appeal to students is that these platforms offer a frictionless friendship. “When a chatbot uses simulated curiosity to ask a student about their day, it signals interest,” the text states. “When it uses social presence to reference something a student shared the week before, it signals memory and care. When it uses sycophancy to affirm a student's feelings without question, it signals safety. These may not be real in the way a human relationship is real—but they can feel real.”

Additionally, talking with a chatbot doesn’t carry a social risk, and these tools are always available and never require anything from a student. In contrast, a real friend might challenge you and disagree with you, isn’t always available and has their own interests and passions. But they do care “about your wellbeing in a real, embodied way.”

4. There Are Signs of Concerning AI Use Educators Can Watch For

The course offers several tangible solutions for addressing concerns about AI use and teaches you some potential signs that a student may have entered into an unhealthy relationship with a chatbot. These include discussing AI as a “real friend,” and sudden switches in academic performance and overall behavior.

In these instances, the course advises not trying to perform the role of a therapist, but instead talking with the student about their AI use in a curious and non-judgmental way, and following your institution’s guidelines around recommending wellbeing and mental health supports. Understanding what students are getting out of their interactions with these machines can also help.

Many students are drawn to chatbots because of their sycophancy (a tendency to agree with the user) and students also have “automation bias,” a strange new tendency many have to trust AI more than their own intuitions because it seems so knowledgeable and confident in its responses.

5. There Are Steps Educators Can Take

When it comes to AI and student wellbeing, as with so many other things, the best interventions are often the ones that take place before a problem develops.

To this end, the course recommends that educators take an active role in preventing students from entering into potentially harmful scenarios with AI. This can be done through digital literacy and by making AI a regular low-stakes topic of conversation in class. Additionally, the course encourages educators to model intentional AI use, explaining why they are using AI and its limitations.

In a great piece of advice for this and other situations as an educator, the course advises teachers not to be afraid to let the students know the limitations of their own knowledge on these topics by saying something such as, “I'm still figuring this out, too.” The course notes that this “is an invitation to think together,” modeling exactly the kind of real human collaboration and thoughtfulness AI can’t duplicate.

Artificial intelligence is quickly moving from novelty to infrastructure. In many schools and universities, AI is already shaping writing, assessment, tutoring, research, student support, and administrative decision-making. UNESCO’s 2026 report, The Algorithm in the Room: Seeing and Confronting the Implications of AI for the Future of Education, argues that educational leaders should resist treating this transformation as inevitable.

The more important leadership question is not simply, “How do we adopt AI?” but “What purposes of education do we want AI to support, and what should remain distinctly human?” In some cases, the broader conversation might also include questions such as, “What is the purpose of our school, program, or course?”

For school and district leaders, higher ed administrators, and instructional technology leaders, the report offers several immediate action items.

1. Start with educational purpose, not the tool

One of the primary principles of effective educational technology has always been to focus on the instructional goal rather than the technology. The UNESCO report repeatedly returns to the problem of alignment. AI systems are designed to optimize toward goals, but schools are complicated, and not all educational goals can be easily quantified in a dashboard or reduced to a simple metric.

The immediate leadership implication is straightforward: before approving a new AI product, ask what educational problem it is intended to solve. Leaders should require a clear instructional or administrative rationale rather than adopting tools simply because of a promise of efficiency or personalization. A useful leadership question is, “What do we want students or educators to be better able to do because this tool exists?”

2. Audit where algorithms are already making decisions

UNESCO’s first recommended horizon is to “see the algorithm.” Educational leaders may be surprised by how many algorithmic systems are already embedded in their institutions, from adaptive learning tools and plagiarism detection systems to enrollment platforms, analytics dashboards, recommendation engines, and automated assessment tools.

An immediate action is to conduct an institutional AI and algorithm inventory. Identify which systems are being used, what data these collect, what decisions they influence, and whether those decisions can be challenged by a human.

This is especially important because UNESCO warns that algorithmic decisions can be opaque, difficult to appeal, and capable of reinforcing existing inequities.

Algorithms need to be tested from time to time. Do not assume that the conditions that allowed an algorithm to be effective a decade ago are still in place.

3. Protect human agency in teaching and learning

One of the report’s strongest warnings concerns the gradual transfer of intellectual work from people to machines. UNESCO describes the possibility of a “classroom AI doom loop,” in which AI generates assignments, students use AI to complete the work, AI evaluates the responses, and automated systems record the grades. The issue goes beyond academic dishonesty and into whether meaningful human learning is taking place.

Leaders should therefore ask instructors to identify the parts of learning that must remain human. These might include forming questions, interpreting evidence, explaining reasoning, engaging in discussion, solving unfamiliar problems, reflecting on mistakes, and defending conclusions. Assessment seems like a key process that needs to remain under human agency. AI can support those processes, but it should not replace any.

4. Rethink assessment instead of relying on AI detection

UNESCO’s discussion of pedagogy and assessment suggests that educational institutions need to move beyond the search for increasingly sophisticated AI detectors.

If AI can produce conventional academic products, leaders should help educators reconsider what those products are intended to demonstrate. Assessment should increasingly make student thinking visible. That could mean more oral explanation, iterative drafts, demonstrations, presentations, authentic projects, reflective commentary, in-class problem solving, and assignments in which students document how they reached a conclusion.

The leadership challenge is to support redesign of assessment to disclose, document, and reflect on AI use by students.

5. Make AI literacy broader than prompt engineering

Many current AI literacy initiatives focus heavily on how to use generative AI effectively. UNESCO suggests that students need something more substantial.

AI literacy should include understanding how algorithmic systems work, how training data influence outputs, how bias can emerge, why hallucinations occur, whose knowledge may be overrepresented or absent, and how commercial incentives shape technology. Students should learn to question AI outputs, not simply generate better ones.

The UNESCO report raises concerns about linguistic and cultural flattening. Dominant languages and highly digitized knowledge traditions may receive disproportionate representation in large language models, while local, indigenous, and less digitized perspectives risk being further marginalized in some cases.

6. Establish clear human-rights and privacy guardrails

UNESCO emphasizes that AI use in education should be grounded in dignity, non-discrimination, participation, equity, informed consent, and protection—especially when children or vulnerable learners are involved.

Educational leaders should immediately review policies governing student data, AI account creation, vendor access, automated decision-making, and the use of personally identifiable information in generative AI platforms.

Institutions also need a clearly articulated guiding principle that consequential decisions affecting students should not be delegated entirely to automated systems.

7. Strengthen teachers rather than designing around them

One of the most important leadership messages in the report is that AI should not become an excuse to hollow out educational institutions. UNESCO argues that hopeful AI futures depend on strengthening libraries, arts and humanities programs, teacher development, and public educational infrastructures.

To achieve this, professional learning should move beyond one-time workshops on individual AI products. Educators need ongoing opportunities to examine AI’s instructional, ethical, cultural, and assessment implications. Teachers should be positioned as designers of learning environments, not simply supervisors of automated systems.

8. Move from consumers to shapers of AI

UNESCO’s final horizon is the most ambitious: “rewrite the algorithm.” The report encourages educational communities to participate actively in shaping AI rather than simply consuming commercially produced systems. It highlights possibilities such as community-trained models, citizen science, indigenous-language projects, locally developed small language models, and collaborative rather than purely individualized AI tools.

For educational leaders, the immediate step may be modest: involve teachers, students, librarians, technology staff, and community members in AI governance and purchasing decisions. Those who experience the consequences of educational technology should have a voice in determining how it is used.

A leadership agenda for right now

Perhaps the most useful takeaway from The Algorithm in the Room is that educational leaders still have agency. AI is not an external force that schools must simply accommodate.

UNESCO’s three-part framework provides a practical leadership sequence: see the algorithm, steer the algorithm, and ultimately help rewrite the algorithm.

For leaders, that means conducting an AI inventory, establishing governance and privacy expectations, supporting assessment redesign, expanding AI literacy, protecting meaningful human interaction, and ensuring that technology remains subordinate to educational purpose.

The most important AI strategy may therefore begin with a question that has very little to do with technology: What kind of education are we trying to provide, strengthen, and create? Once leaders can answer that clearly, decisions about AI become considerably easier.

Of course, what kind of education we should be providing might be the most difficult question of all.

Schools across the nation are being told to limit screen time. That sounds reasonable until you ask a basic question: What exactly are we limiting?

A student could spend 30 minutes passively clicking through a poorly designed program. Another could spend 30 minutes researching a topic, reading, collaborating with classmates, writing, or creating a podcast or video. Same 30 minutes. Very different learning.

It's a little like calories -- 300 calories of avocado is not the same as 300 calories of Twinkies. The number tells you how much, but not enough about the quality.

Screen time has the same problem. A minute watching random videos is not the same as a minute reading, writing, creating, collaborating, coding, or getting feedback. In my previous Tech & Learning article, Screen Time Is the Wrong Unit of Measurement, I argued that instead of measuring minutes, we should be measuring learning.

Interestingly, MIT is taking a similar approach in its new report on AI and education. Rather than simply embracing or rejecting AI, MIT's Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training recommends "AI-aware" education. That means revisiting what students need to learn, changing assessments, strengthening hands-on learning, maintaining human interaction, and helping students understand when AI should and shouldn't be used.

That's the conversation schools should be having about technology more broadly. Don't start with the tool or the number of minutes. Start with the learning and decide what tools are best to achieve the desired outcomes.

We should also be helping students think critically about their own technology use. They need to recognize when technology helps them learn, create, and connect, and when it distracts, gets in the way, or results in cognitive offloading. Counting minutes doesn't teach that judgment, and with technology evolving constantly, judgment matters more than a timer.

Measure What Matters

Instead of counting minutes, we should evaluate the quality of the learning experience. Other considerations exist, too. Privacy and safety should be prerequisites for any technology we put in front of students.

Here are some of the areas we should be looking at:

Learning value and engagement: What is technology adding? Are students learning, thinking, creating, questioning, solving, and making decisions, or mostly clicking and consuming?

Accessibility: Is the technology helping the student by making the learning more accessible through tools such as text-to-speech, speech-to-text, captions, translation, audio, enlarged text, or other supports?

Personalization and agency: Does technology let students learn at an appropriate level or pace and make meaningful choices about how they learn, create, communicate, or demonstrate what they know?

Human interaction: When interaction matters, is technology helping students connect and collaborate, or getting in the way?

Transformation: Does technology simply replace an analog tool, or does it change what students can do? The SAMR model gives us a useful way to think about this through substitution, augmentation, modification, and redefinition. Not every activity needs to reach redefinition, but we should be able to move beyond substitution and explain what the technology adds.

Evidence and outcomes: Did it work? What evidence shows that students learned, gained access, improved their work, or achieved the intended outcome?

A screen-time limit is a lazy measure. Pick a number, start a timer, and you're done. We've done this before with multiple-choice tests. These are easy to score, but easy to measure doesn't mean meaningful. We should be moving toward better ways of evaluating learning, not backward toward whatever is easiest to count.

Empower Students to Know When the Screen Should Go Away & When It Should Stay

None of this means students should be on devices all day. Support students in determining if technology is supporting their learning. If technology isn't adding anything, they shouldn’t use it. If students would learn more with pencil and paper, by moving, building, experimenting, performing, playing, or working with their hands, let them.

But the reverse is also true.

If a student needs text-to-speech or translation to access a book, let them use the screen. If they can collaborate with experts or peers they couldn’t reach without technology, let them. If they can authentically publish their work, empower them with a screen. Students also shouldn't be called out, stigmatized, or othered for learning through different modalities. The modality should not matter if the student is learning.

The question shouldn't be, "How much screen time did students have?" It should be, "What were students doing, what were they learning, and was technology helping?"

Generative AI has created a new challenge for teachers. Students can ask tools such as ChatGPT, Claude, and Gemini almost any question and receive a polished answer in seconds. The problem is that polished does not always mean accurate.

Large language models (LLMs) can produce incorrect information, misrepresent sources, and even invent convincing citations to publications that do not exist, which is why many educators remain understandably cautious about students using AI for research.

The better response is not to tell students to avoid AI altogether but to teach them how to use AI tools that keep evidence at the center of the research process.

What Is A RAG?

A retrieval-augmented generation tool, or RAG, helps students move from asking AI for answers to asking AI to help them find, examine, and explain evidence.

One way to explain the difference between a traditional LLM and RAG is to compare it to the difference between a closed-book and open-book exam. When students use an LLM by themselves, the model answers from what it learned during training and from whatever the student provides in the conversation. It may produce a useful response, but it may also fill gaps with something that sounds right rather than something that is right.

RAG changes that interaction. Before answering, the system retrieves information related to the question and gives that material to the model. In that sense, RAG is less like asking AI to guess and more like asking AI to work with evidence in front of it.

The process is essentially: Question → Retrieve sources → Analyze sources → Generate response

This distinction is important for students because it introduces something often missing from conventional AI interactions: an evidence trail.

RAG does not eliminate hallucinations, and students still need to verify AI-generated information. However, properly designed retrieval systems make verification much easier because students can trace claims back to underlying sources.

Sometimes the term agentic RAG is used; most current RAG tools are better described as RAG with agentic features, not fully agentic RAGs at this point. These retrieve evidence, but may not independently plan, evaluate, and revise the research process in the way a true agentic system would.

With that foundation in mind, here are five practical ways educators can begin using retrieval-based AI tools to help students ask better questions, locate stronger evidence, and verify what AI tells them.

1. Use Gemini Notebook to Create a Controlled Research Environment

Google's Gemini Notebook provides perhaps the easiest introduction to RAG for students. Rather than relying primarily on the model's general knowledge, teachers and students can provide the sources Gemini Notebook will use. These can include PDFs, websites, Google Docs and Slides, YouTube videos, and other materials. Responses are grounded in those sources and include citations students can follow back to the original material.

That allows teachers to create a controlled AI research environment. For example, instead of asking students to use AI to investigate the causes of the American Revolution from the open web, a teacher could provide six carefully selected primary and secondary sources. Students could load those materials into Gemini Notebook and ask: "What were the most important colonial objections to British taxation?"

The assignment also shouldn't end with the answer. Require students to identify the sources Gemini Notebook used, locate the supporting passages, and determine whether those passages support its conclusions.

Suddenly, AI isn't replacing reading but giving students another reason to read critically.

2. Use Consensus to Move Students Toward Scholarly Evidence

For older students, particularly those beginning serious research projects, Consensus can provide a useful bridge from general web searching to scholarly literature. Instead of asking a general-purpose chatbot to recall studies from its training, students can enter natural-language research questions and receive results grounded in academic research. Consider the difference between asking: "Does homework improve student achievement?" and asking: "What peer-reviewed research suggests that homework improves student achievement?"

The second question establishes an evidentiary expectation. Students can then examine the research Consensus retrieves, determine which studies address their question, and follow those studies to the original publications.

The objective is to help them identify research worth investigating rather than have Consensus do it for them.

3. Use Elicit for More Sophisticated Research Projects

Elicit takes this process another step and is particularly useful for advanced secondary and postsecondary research. Students can use it to locate scholarly literature, compare studies, extract information, and begin organizing evidence for literature reviews. This provides an opportunity to teach something students frequently struggle with finding a source is not the same as evaluating a source.

A student might use Elicit to identify ten studies related to a research question. The next task should be determining which studies belong in the research project. Students can examine methodology, sample size, population, publication date, limitations, and relevance to the research question.

In this example, AI accelerates discovery but students remain responsible for judgment. That distinction should become central to AI literacy.

4. Use Scite to Ask, "Does Research Really Support This?"

Students often assume that finding a citation proves a claim. Scite provides a useful way to challenge that assumption. Its Smart Citations show how later scholarly publications have cited earlier work and classify citation statements as supporting, contrasting with, or simply mentioning the original study.

Imagine a student finding a 2018 study that makes an important claim. Rather than simply citing it, the student can ask better questions: What happened to this finding afterward? Have later studies supported it? Challenged it? Qualified it? Has the scholarly conversation moved in another direction?

This approach introduces students to an important feature of authentic scholarship: knowledge isn't simply a collection of isolated facts but an ongoing conversation in which evidence is continually examined and reconsidered.

5. Use Perplexity for Current Information, but Verify the Sources

Not every student research question belongs in an academic database. Students researching current legislation, technology developments, elections, economic conditions, or breaking events need access to information that changes quickly.

Retrieval-oriented tools such as Perplexity can be useful because it searches current web information and connects generated answers to sources. However, this presents another excellent teaching opportunity. A citation does not automatically make information credible. Students should examine who published the information, when it was published, what evidence the source provides, whether other credible sources corroborate it, and whether the AI accurately represented what the source said.

The presence of citations should therefore be the beginning of verification, not the end of it.

Establish One Simple Classroom Rule

Teachers introducing RAG do not need an elaborate AI policy to get started. Begin with one simple rule: Never cite the AI tool, but cite the evidence behind the AI.

When an AI tool provides a source, students should open it. When it makes a factual claim, students should find supporting evidence. When it summarizes research, students should compare the summary with the original. And when the evidence is not strong enough, students should keep researching.

Establishing these habits turns AI use from passive answer collection into active information evaluation.

Create an AI Research Toolbox

Students also need to understand that there isn't one best AI tool. Most of these tools have clear strengths and weaknesses, so they will have to learn to use the one best for each type of question or problem.

• Assigned readings and teacher-selected sources: Gemini Notebook
• Finding scholarly evidence: Consensus
• Conducting deeper literature reviews: Elicit
• Determining whether research supports or challenges a claim: Scite
• Investigating current information: Perplexity
• Brainstorming, tutoring, explaining, organizing, and feedback: General-purpose LLMs such as ChatGPT, Claude, or Gemini

This shift represents an important evolution in instructor and student AI literacy. Instead of teaching students simply how to prompt an AI system, we need to teach them which information environment they should ask the AI to search.

Getting Started Tomorrow

Teachers can introduce this approach without redesigning an entire course. For example, choose three to five credible sources related to something students are already studying and put them into Gemini Notebook. Give students a research question and ask them to generate an initial response. Then add the requirement that students verify three claims in the AI response against the original sources. For each claim, students should identify the source, locate the supporting evidence, and decide whether the AI represented that evidence accurately. A follow-up discussion can focus on what the AI got right, what it missed, and where its interpretation differed from students' interpretations.

That's a relatively small instructional change, but it introduces several important competencies simultaneously: AI literacy, source evaluation, corroboration, close reading, and evidence-based reasoning.

Move From Prompt Literacy to Evidence Literacy

The first phase of generative AI in education focused heavily on prompts. Students and teachers learned that better questions could produce better AI responses.

The next phase needs to go further. Students need to understand where AI-generated information comes from, why some sources deserve more credibility than others, how retrieval influences an AI's response, and how to determine whether the evidence supports what the AI says.

Ultimately, the most important question we ask students may no longer be, "Did you use AI?" A much better question is, "What evidence did the AI find, and how did you determine that evidence was trustworthy?" This then becomes information literacy for a world in which students will increasingly encounter information through AI systems.

Throughout my work with educators and district leadership teams across the country, I have seen growing interest in and a need for AI professional learning. Educators want to understand what AI is, how it works, and how it can support their work, but they also want practical guidance about data privacy, academic integrity, bias, responsible classroom use, and appropriate student use. They need clear and consistent guidance that supports their practice.

A single workshop or training will not adequately prepare educators for everything they need to know and do. I have been passionate about this work for the past four years since starting my doctoral research, which focused on the need for effective AI professional development for educators.

A key takeaway from my research was the importance of ongoing, relevant, and supportive professional learning. My work with schools and educators has continued to reinforce that finding.

Effective AI professional learning cannot simply be about showing teachers the latest tools. Tools change, policies evolve, and new concerns emerge. If PD focuses primarily on a platform, what educators learn may become outdated before they have an opportunity to apply it in their practice. We need to build knowledge, confidence, and skills that educators can transfer regardless of what AI tools are available.

Start With Purpose, Not With Tools

For years, when I provided PD, I often focused on the tools. Over time, I recognized that technologies change, and what matters most is understanding the why behind using technology. The same is true with AI. Begin by asking why an educator might use AI rather than asking which AI tool they will use.

When I work with educators, I challenge them to identify specific needs in their own practice. Do they want to differentiate instruction? Create more accessible materials? Provide feedback? Redesign an assessment? Generate ideas? Reduce time spent on repetitive tasks? Starting with an instructional need keeps the focus on teaching and learning rather than technology for technology’s sake.

Providing guiding questions helps educators decide when AI is useful and when it is not. Effective professional learning should guide teachers to ask:

• What is the learning goal?
• What thinking should students do independently?
• Where might AI provide meaningful support?
• What would be gained or lost by using AI?
• What happens if AI completes too much of the work?

These questions will remain relevant even as the technology changes.

Connect Learning to Real Classroom Practice

Just as relevance makes learning more meaningful for students, professional learning needs to connect to educators’ actual roles and responsibilities. A kindergarten teacher, high school science teacher, school counselor, special education teacher, instructional coach, and district administrator may ultimately use the same AI platform, but they will not necessarily use it in the same way.

Professional learning should include realistic scenarios for grade levels and content areas, and provide an opportunity for participants to create something they can use right away. Teachers might revise an assignment, develop expectations for student AI use, evaluate an AI-generated response, design a differentiated resource, or create a lesson that requires students to verify AI output.

In my sessions, educators have found it valuable to work through scenarios with colleagues rather than just learn what AI can do. Some scenarios I use include a student using an unapproved chatbot, evaluating a biased AI-generated response, considering the consequences of exposed student information, or addressing suspected AI-assisted student work.

Discussions enable educators to collaborate and think about what they would do, consider the benefits and cautions of AI, and learn together. These conversations help them leave with a clearer understanding of AI and specific next steps for their own learning and practice.

Differentiate the Professional Learning

We differentiate learning for students, so we should think similarly about professional learning for educators. Everyone needs a solid foundation in AI literacy, but not everyone needs the exact same training.

Effective professional learning should provide opportunities for educators to explore how AI connects specifically to their work. Classroom teachers may need more support with instructional design, assessment, AI literacy, and appropriate student use. Administrators may focus more on leadership, policy, implementation, and communication. Technology teams may need more time on data privacy, security, and vendor vetting.

Professional learning should also account for differences in experience and comfort with AI. Two teachers in the same department may have very different starting points. One may already be experimenting with AI while another is still learning the basics. Providing opportunities for choice and time to learn with colleagues meets educators where they are without overwhelming them.

Address Privacy, Law, and Ethics

Responsible AI use cannot be an add-on at the end of professional learning. Privacy, safety, ethics, accessibility, and legal considerations need to be woven throughout the experience.

As an educator and attorney, I frequently work with schools that have questions about student data, approved tools, vendor agreements, copyright, accessibility, and academic integrity. Many educators have expressed not having enough guidance about what information can be entered into an AI platform or how to determine whether a tool is appropriate for students.

Educators need practical guidance rather than a list of laws without classroom context. Questions to consider:

• Why does entering identifiable student information into an unapproved platform create risk?
• What do age requirements mean for student access?
• What should educators look for in privacy policies?
• Why must AI-generated content be reviewed for bias and accuracy?

Schools need consistent expectations. Professional learning should connect directly to district policies and show what those policies mean in practice.

Make Professional Learning Active and Ongoing

In schools, a keynote or introductory workshop helps to build awareness and momentum, but it should serve as the beginning of the learning process rather than the entire strategy. Ongoing support might include workshops, content- or grade-level coaching, professional learning communities (PLCs), office hours, model lessons, and ongoing conversations.

The goal is not to become an AI expert. Educators need enough understanding and confidence to make informed decisions, recognize when AI adds value and when it does not, and guide students to do the same. When this is possible, we move from simple awareness of AI and its possibilities to intentional implementation.

Include Students and Families

AI professional learning should help educators understand how students are using these technologies. Students may be employing AI to brainstorm, translate, summarize, study, create images, or complete assignments. Some students may understand the limitations of these tools, while others may trust the output without questioning it.

Ask students how they are using AI. What do they think is permitted? What do they consider cheating? When do they believe AI helps them learn? When might they be relying on it too much? What do they know about the information they are sharing with AI tools? Student voice can reveal gaps between school policies and actual classroom practices. It can also help schools develop expectations that are realistic, understandable, and responsive to student needs.

Families should be part of the conversation as well. Parents and caregivers may have questions about why AI is being used, what tools students can access, how their children's information is protected, how schools define appropriate use, or how AI changes the expectations for student work. Professional learning should prepare educators to answer these questions.

Build Systems That Can Adapt

Effective AI professional learning should be part of a larger system that involves leadership, policy, curriculum, technology, families, and students. District leaders must communicate a clear vision, provide guidance about approved tools, and reflect on decisions as the technology changes. Schools should also evaluate whether professional learning is making a difference in educators’ confidence, classroom decisions, assessment practices, and understanding of privacy expectations.

The measure of successful AI professional learning goes beyond the number of tools educators explore during a workshop; it is what happens after the workshop. Are educators asking better questions? Are they protecting student information? Are they making intentional choices about when AI adds value?

The tools in use today may not be the tools used next year, which is why understanding the fundamentals of AI and how it enhances learning is essential before diving into specific tools. Effective professional learning is not an experience that tells educators to embrace every new tool or avoid AI completely. The best learning helps educators understand AI, its capabilities, and guides them to approach it with curiosity, caution, and confidence, with the human side of education at the center.

As someone who writes and speaks frequently about artificial intelligence in education, I consider myself a strong proponent of its thoughtful use in teaching, learning, and school leadership. Because AI has become such a regular part of my own professional work, I was genuinely surprised to learn how little my daughter, a freshman in high school, had used it, and more importantly, how skeptical she was about its growing role in school.

Her concerns grew largely from her identity as an artist. She spends a great deal of time drawing and creating, so she is especially attentive to the ways AI-generated art may affect human artists, particularly when original work can be copied, repurposed, or used without the creator’s knowledge. What began as a conversation about AI and art quickly expanded into a broader discussion about effort, learning, creativity, and the growing dependence on AI among students.

I was struck by how mature and carefully considered her views were, especially because they differed so sharply from my own generally positive stance toward AI. Rather than trying to persuade her or correct her assumptions, I became curious about how she had arrived at those conclusions and what she was observing in school.

That curiosity led me to ask whether she would sit down with me for a brief interview. I emphasize “brief” because, like many teenagers, she has a limited tolerance for anything that feels even remotely school-related, especially when the interviewer happens to be her father.

During our conversation, I asked her what AI means to her, what she is seeing among her classmates, and whether she believes it has a constructive role in education. Her answers were candid, skeptical, and often surprisingly nuanced. While adults continue to debate what AI may eventually mean for schools, she described what it already looks like from the perspective of a student living through that change.

This article is based on a semi-structured interview with my daughter and her perspective is not intended to represent the experiences or opinions of all students. Qualitative conversations such as this can provide meaningful insight into how individual students are making sense of emerging technologies in their schools and daily lives.

AI as a Substitute for Thinking

I began by asking what came to mind when she heard the phrase “artificial intelligence.” Her answer was immediate:

“A robot who does work for you because you can’t do it yourself.”

When I asked her to explain, she continued:

“You just type in a prompt and it does it for you instead of using your own brain and doing the work yourself.”

Her description was not quite how I would define AI, and part of me wanted to explain that a large language model does not literally think or work in the same way a person does. I intentionally resisted that urge because I wanted to understand her perspective rather than shape it. I was interested in how a freshman interpreted AI through her own experiences, not whether she could provide a technically precise definition.

Even so, her response was striking because she did not begin with the language adults often use when discussing AI. She did not describe it as innovative, efficient, exciting, or transformative. Instead, she framed it around effort and ownership. In her view, AI becomes problematic when it replaces the thinking a person should be doing. That distinction would continue throughout the interview. She was not simply asking whether AI could produce an answer, but what happens to the learner when the tool takes over the work.

Seeing Dependency Among Students

Although she said she does not really use AI herself, she described seeing it frequently among classmates:

“A lot of students in my class use it to their own advantage, saying, ‘Oh, I don’t know this question. I’m just going to use AI to do it.’”

One of the clearest examples came when her school removed student access to Google Gemini. She recalled that students immediately reacted:

“The kids started freaking out.”

When I asked why, she answered:

“Because they use AI so often, they depend on it.”

She also described the kind of response she heard from peers:

“What am I going to do? I can’t turn in my essay if I don’t have it.”

That moment captures more than frustration over losing access to a tool. It suggests that, for some students, AI may already have become embedded in the process of completing schoolwork. When the tool disappears, so does their confidence in being able to produce the work independently.

Her concern was not simply that students were using AI, but that some were becoming unable, or at least unwilling, to proceed without it.

When Grades Matter More Than Learning

Perhaps the most mature observation she made came when she explained why students might turn to AI so quickly. Rather than blaming students alone, she connected their behavior to the incentives surrounding them:

“I think a lot of the world has made getting grades more important than learning the actual stuff.”

She then put the issue even more directly:

“Grades matter more than what they actually get in their brain.”

That statement deserves serious consideration from educators. It suggests that AI use cannot be separated from the larger culture of schooling. When students believe that the final grade matters more than the process of understanding, tools that produce faster answers become especially appealing.

In that sense, AI may not be creating an entirely new problem but intensifying an old one. Schools have long struggled with the tension between performance and learning, and now AI makes it easier for students to produce polished work without necessarily developing the knowledge behind it.

Her comments also challenge educators to look beyond detection and enforcement. If students are using AI because they feel pressure to produce, then the conversation must include assessment design, grading practices, workload, and the messages schools send about what success actually means.

AI as a Tutor, Not a Replacement

Despite her skepticism, she did not reject AI completely. Instead, she drew a clear line between using AI to learn and using AI to avoid learning.

When discussing a difficult math problem, she said:

“Instead of having AI give you the answer, maybe use it properly and have it create a good step-by-step way to help you.”

She then summarized her view in a single sentence:

“It should be a tutor instead of someone to do the work for you.”

This may be the central idea of the entire interview. For her, the value of AI depends on the role it plays. A tutor explains, prompts, models, and supports. A replacement takes over the task and removes the learner from the process.

That difference is especially important for schools because it offers a more useful framework than simply labeling AI as good or bad. The better question is whether the student remains intellectually engaged. Is the tool helping the student understand, or is it allowing the student to bypass the work entirely?

Her answer was consistent: AI should help people think, not think in their place.

A Human-First View of Art

Her perspective on AI is deeply shaped by the fact that she is an artist. Drawing is not merely an activity she enjoys. It is a skill she has spent time developing, and she sees creativity as something connected to practice, effort, and personal expression.

When I asked why she had not used AI very much, she explained:

“People who use AI to make art are looked down upon, in my opinion, because they can’t draw it themselves or they have to use a robot to generate it.”

Her language was blunt, but the underlying concern was about authenticity. She values the work that goes into developing an idea, choosing how to represent it, making mistakes, revising, and improving. From her perspective, generating a finished image through a prompt can bypass the very process that gives artistic work meaning.

At the same time, she offered a more constructive possibility:

“Instead of using a bot to generate a piece of art, you can draw something yourself and then ask AI, ‘How can I make this art piece better?’”

Once again, she returned to the distinction between support and substitution. She was more open to AI when the human remained the creator and used the tool to receive feedback. She was far less comfortable when AI became the creator itself.

Her concern also reflects a broader anxiety among artists about how creative work may be copied, repurposed, or used without permission. For a young person developing her artistic identity, those concerns are not abstract but directly affect how she thinks about ownership, originality, and whether technology is respecting or erasing human contribution.

Schools May Be Responding Too Late

When I asked what schools should teach students about responsible AI use, she raised another important issue: timing.

“AI has already come out, and teachers are a bit late teaching it.”

Her point was that students have already encountered and formed habits around AI before many schools have developed clear guidance. By the time formal instruction arrives, some students may already see AI as a normal way to complete assignments.

She was also skeptical that instruction alone would change behavior:

“I don’t know if the kids would care about how they teach it. They’re still going to use it irresponsibly.”

Her response may sound pessimistic, but it reflects a practical understanding of student behavior. A single lesson on responsible use is unlikely to undo habits that have already formed, especially when those habits are reinforced by convenience, pressure, and easy access.

This suggests that schools need more than one-time AI literacy lessons. They need consistent expectations, well-designed assignments, clear boundaries, and conversations that help students understand why certain uses support learning while others undermine it.

The Role of Teachers

When I asked what role teachers should play as AI becomes more common, she was not fully convinced that AI should become a regular part of education. Still, she believed that if teachers use it, they should define its role carefully:

“They can say, ‘This is going to be a secondary tool or a helper. You can only ask questions to help you instead of giving you the answer.’”

Her use of the phrase “secondary tool” is important. She does not believe AI should sit at the center of learning. The teacher, the student, and the learning task should remain primary, while AI occupies a limited and clearly defined supporting role.

That view also places responsibility on educators. Students need more than access to a tool. They need examples of appropriate use, boundaries around inappropriate use, and opportunities to practice using AI in ways that deepen rather than replace thinking.

Her Message to School Leaders

I ended by asking what message she would give to teachers and principals.

Her response was brief and direct:

“Please don’t generate art pieces, and please have better restrictions on AI usage.”

The first part reflects her identity as an artist. The second reflects her concern that current expectations are not strong or clear enough.

Her answer also complicates the assumption that students always want fewer restrictions. Some students may actually be looking to adults to set firmer boundaries, especially when they see classmates becoming dependent on tools that may weaken their learning.

What Her Perspective Taught Me

As I reflected on the interview, I realized that her comments went beyond technology, and were about effort, creativity, integrity, learning, and what students believe school is asking them to value.

Her perspective may not represent every student, and it should not be treated as though it does. Some students are enthusiastic about AI, some are cautious, and others are still deciding what they think. Her voice matters precisely because it adds complexity to a conversation that is often framed too simply.

Perhaps what surprised me most was not her skepticism but how little she had actually used AI. As someone who advocates for the thoughtful integration of AI in education, I expected that a high school freshman would already be experimenting with these tools regularly. Instead, I found myself listening to a student who had largely formed her opinions by observing the behaviors of her classmates rather than relying on AI herself. That realization reminded me that student perspectives on AI are far more diverse than we sometimes assume.

She is skeptical, but she is not closed-minded. She objects to AI when it replaces thinking, creativity, or human effort, yet she can still imagine a constructive role for it as a tutor, guide, or source of feedback.

The consistency of her view was what stood out most. AI should explain, support, and improve. It should not take over.

Perhaps the most important lesson from our conversation is that schools should not design AI policies based solely on adult assumptions about what students need. Students are already observing how AI is affecting their peers, their classrooms, and their own sense of what counts as authentic work. We should listen to them more carefully because they are not merely preparing for an AI-shaped future. They are already living in it.

I was not prepared for just how much I would hate using Speaker Coach, but that doesn’t mean it wasn’t helpful.

The AI-powered speaking coach and presentation mentor built into PowerPoint is designed to help anyone, including educators and their students, improve their public speaking abilities.

I’ve always been comfortable with public speaking, and think I do a solid job. My voice carries, I speak confidently, and I feel I know how to read a crowd and respond in the moment to their moods. But I am by no means perfect. I know that I can mumble, I lose my train of thought sometimes, and when I read from a prewritten script or speech, I have a tendency to read it in a semi-lifeless robotic voice. Also, I have a bad habit of reading too much from slides, especially when I get nervous, which is why I often avoid using slides in my presentations.

Speaker Coach caught on to all these weaknesses, almost immediately, which was, frankly, a little bit disheartening. I always hoped I was being too hard on myself and that I was actually far better at public speaking than I thought. This made using Speaker Coach an unpleasant experience, but it also meant it was doing its job and could help me improve my presentations.

So overall, despite some minor issues, Speaker Coach seems to be a strong, if slightly intimidating, tool for improving public speaking. Here’s a closer look.

How Speaker Coach In PowerPoint Works

Speaker Coach is built into PowerPoint. To access it, users access the web version of PowerPoint through a Microsoft-linked account. They then upload an existing slideshow or create a new one.

Once a user is ready to practice, they select the Slideshow tab, then choose the Rehearse with Coach option. After this, you can move through your presentation as you talk and practice with Speaker Coach monitoring it all.

As you speak, the program gives on-screen guidance by default, though you can turn off this feature. It will provide feedback such as telling you to speak more slowly, avoiding “ahs” and “ehs,” and offering encouragement that you are doing great and should keep it up.

Once you finish the presentation, you get a breakdown of your overall performance, including a wide range of useful feedback on your presentation. It will let you know, for instance, how fast you spoke and it was too rapid or too slow, and whether you read from the slides directly too often, and if so, which slides. It also offers a chart on how your pronunciation was, as well as your tone variation and your use of inclusive language, as well as many helpful details such as how long you spent on each slide.

What Are Speaker Coach’s Best Features?

Public speaking is often a vital skill for educators, but it can be a tricky one to practice. Speaker Coach offers the chance for hands-on practice with specific and actionable feedback.

The final report is really helpful in assessing your strengths and weaknesses as an instructor. The fact that it recognized the slides I had a tendency to read too much from was fantastic.

It can also help you identify the strengths and weaknesses of different types of presentations. I practiced with scripted presentations and with unscripted ones in which I relied only on my slides. I learned, unfortunately but not particularly surprisingly, that I have too many "ums" and "ahs" when I don’t use a script. Reading from a script corrects this, but then my voice becomes more lifeless and monotone. Identifying this helps me target both concerns as I prepare for future presentations.

What Speaker Coach Features Need Work?

As I mentioned at the start, despite finding it helpful, I didn’t enjoy using Speaker Coach. Some of this is probably inherent in practicing a skill as deeply personal as speaking, but there were also design features that made this worse.

The real-time feedback tool was more annoying than helpful, and I was tempted to turn it off, which is option. If I said “ah" or "mmm” it would start scolding me, which made me feel like a kid when adults used to admonish me to "stop mumbling." Even when it praised what I was doing, it was distracting.

Another weakness of Speaker Coach overall is that even though it was great at diagnosing the problems in speech, its suggestions for fixing these problems are limited.

It told me, for instance, to avoid saying “ah” by pausing and breathing, which I know I should do, except it didn’t offer more concrete advice on actually practicing this skill or how I might build in these natural pauses. When it came to speaking in a monotone, Speaker Coach didn’t offer any suggestions of any kind for improvement.

Overall, more detailed suggestions and/or links to resources on public speaking would take this tool to the next level.

How Much Does Speaker Coach Cost?

Speaker Coach is free with PowerPoint, which comes with most Microsoft Office plans available through schools or businesses.

Individual subscriptions range between $6.99 and $9.99 per month, depending on whether users pay annually or monthly.

Bottom Line on Speaker Coach

Ultimately, I found Speaker Coach helpful, if unpleasant, to use, but I plan on using it in the future as I try to improve my public speaking ability. I would recommend it to educators and students who are already comfortable with public speaking and looking to really hone their craft.

I wouldn’t recommend it to your average student who is working on a presentation for class because I worry the no-mercy criticism Speaker Coach provides could have a negative effect on someone who is new to presentations and likely already nervous about public speaking.

However, for the right type of educator or student with thick skin, Speaker Coach can be a helpful speaking coach.

Sometimes teachers need a quick, useful resource that can slot into a lesson without hours of preparation, which is exactly what Rumie is designed to provide.

Rumie Learn is a nonprofit learning platform built around short, interactive "Bytes," with more than 2,500 published learning experiences covering everything from digital literacy and career skills to health, financial planning, and learning strategies. The platform has reached more than 2.5 million learners, with new Bytes added regularly.

This doesn't use lengthy courses, but instead breaks down subjects into small chunks that can generally be completed in around five minutes. That makes it great for teachers looking for starters, discussion prompts, flipped-learning activities, or quick ways to introduce a new topic.

What is Rumie?

Rumie is a free online library of microlearning content designed to make useful knowledge easier to consume and remember.

The basic unit is the Byte. These are short learning experiences, generally under 10 minutes, built around a specific concept or practical skill rather than an entire subject. Rumie says its Bytes use a chunked structure intended to make learning more efficient, while its lessons commonly combine short pieces of text with images, GIFs, memes, and quizzes.

That format makes Rumie rather different from a traditional learning management system. It doesn't have a huge course to enroll students in and work through from beginning to end. Instead, teachers can search the library for a specific idea and use an individual Byte when it makes sense for their lesson.

The library covers categories including Career Skills, Digital Literacies, Civics, Health, Business Acumen, Learning to Learn, and Innovation. More specific topic tags are also available, making it easier to find resources around particular skills or learning approaches.

Importantly for schools, the content isn't restricted to traditional academic subjects. A teacher could use one Byte to introduce digital citizenship, another to explore critical thinking, and another to help students understand financial planning.

That gives Rumie a useful role alongside existing classroom resources rather than trying to replace them.

How does Rumie work?

The simplest way to use Rumie Learn is to search its library, find a relevant Byte and share it with students.

Rumie also has guidance specifically for teachers on using Bytes as part of a lesson. Its recommended approach is to choose a Byte based on the skill being developed, share it with students, and then use the content as a starting point for discussion and application. That means a Byte doesn't have to be the lesson. It can be the thing that gets the lesson moving.

For example, students could complete a five-minute Byte before class as a flipped-learning activity. The teacher can then spend classroom time discussing what students learned, challenging their understanding, and asking how they could apply the idea.

Alternatively, the whole class can work through the Byte together. Rumie's own lesson-planning example suggests reading a Byte as a group or having students work in pairs before discussing their thoughts as a class.

This is where the short format becomes useful. A teacher can introduce a new concept without committing an entire lesson to a digital course, leaving plenty of time for explanation, debate, practical work, or assessment.

The platform is also designed around accessibility of learning rather than assuming students have hours available. Rumie describes its Bytes as quick, commitment-free learning experiences that can build toward broader skills over time.

What are the best Rumie features?

Thousands of short learning resources
The biggest advantage is simply the size and breadth of the library. Rumie currently lists more than 2,500 published Bytes, while its homepage highlights topics ranging from career skills and digital literacy to health, civics, and learning strategies.

New Bytes are published regularly, giving teachers an expanding pool of material from which to draw. That makes the platform useful when you need something specific rather than a complete curriculum.

Microlearning that fits into existing lessons
A five-minute resource is much easier to fit into a normal lesson than a 40-minute online course. A Byte can become a starter activity, homework task, revision exercise, discussion prompt, or introduction to a new topic. Teachers can also combine several Bytes around a broader theme.

Rumie's own classroom guidance suggests using Bytes in sessions ranging from 30 to 60 minutes, with time built in for discussion and applying the learning.

Interactive content
Rumie uses visual content, GIFs, memes, quizzes, and other interactive elements to make short lessons feel less like reading a textbook. The format is designed to keep the learner moving through a single idea rather than presenting a large amount of information at once.

Content created by professionals
Bytes can be produced by corporate employees taking part in volunteer workshops, community partners, and Rumie's own learning designers. Rumie says its learning designers plan, author, and peer-review Bytes, while the organization vets content against its standards.

Each Byte also identifies its author and, when applicable, the organization behind the content. That provides teachers with some useful context about from where a particular learning resource came.

Useful beyond traditional subjects
Rumie can be valuable for the skills that don't always fit neatly into a timetable. Its library includes resources covering areas such as communication, career planning, financial skills, digital citizenship, mental health, and learning strategies.

Classroom-specific resources are available, covering topics such as inclusive classroom management, facilitation techniques, and accessible worksheets. That makes it a potential source of material for PSHE, pastoral learning, careers education, digital citizenship, and tutor time as well as more conventional lessons.

How much does Rumie cost?

Rumie is free to use. That's one of its biggest attractions for teachers and schools. Rumie is a nonprofit organization whose stated mission is to make learning accessible, and its public Learn platform provides access to its library of Bytes without charging learners for the individual resources.

Separate Rumie products are aimed at organizations, including Rumie Connect and Rumie Build, which allow for the curation of learning experiences or the ability create their own microlearning content. Those are part of Rumie's broader learning platform rather than requirements for using the public Rumie Learn library.

For an individual teacher looking for ready-made learning material, however, there's no subscription to pay for simply browsing and using the public Bytes.

Rumie Tips

• Use Bytes as lesson starters
  Don't feel you need to build an entire lesson around Rumie. Find a five-minute Byte that introduces the concept you want to teach, then use it as the springboard for your own explanation or activity.
• Try flipped learning
  Send a Byte to students before a lesson and use the classroom time to discuss it. This allows students to encounter the material before they arrive and gives teachers more time for questions and application.
• Make students explain the learning
  After completing a Byte, ask students what surprised them, what they found useful and how they could apply the idea. This turns a short digital activity into a discussion rather than another piece of passive screen time.
• Build a themed sequence
  Search for several Bytes around a common subject and use them across multiple lessons. Because each resource focuses on a relatively small concept, teachers can gradually build a broader learning journey without asking students to complete one enormous course.
• Let students explore independently
  Rumie's huge range also makes it useful for independent learning. Ask students to choose a Byte on a subject they know little about, complete it and then present the most useful thing they learned to the class.

The pace of advancement in artificial intelligence tools is breathtaking. Since the public release of ChatGPT in November 2022, multiple platforms have entered the AI arena, all vying to win adherents to their particular flavor of AI. While students may welcome the ability to quickly get answers without much work, educators and education leaders have been deeply uneasy about the potential effects of AI on learning.

With the growth of these platforms comes some desirable changes for education. One of the greatest concerns from the beginning has been its potential to disrupt genuine learning in favor of the nearly-instant answers and solutions provided by AI tools. The good news is that nearly all of the tools here are structured (in theory at least) to support true independent learning and critical thinking rather than serve as answer machines.

To learn more, continue reading — and be sure to read the full Tech & Learning reviews available for most of the listed platforms.

Best AI-Powered Tutors for Education

Gemini’s Guided Learning Mode
Rated highly by Tech & Learning writer and educator Erik Ofgang, Gemini’s Guided Learning Mode is built on the AI model LearnLM, which is designed for education. Erik tested the tool on a variety of topics, including wave equations in quantum mechanics and journalism (a subject he teaches at the post secondary level). While it isn’t perfect, Erik concluded that it’s currently “the best example I’ve seen of what an AI tutor can be.” Free.

Khanmigo
Top-rated by writers from Common Sense Media, the Wall Street Journal, and The Washington Post, Khanmigo is the AI-powered tutor from the well respected education nonprofit Khan Academy. It provides test prep, feedback on writing and facilitates deep learning—without giving answers. In addition to its role as a tutor for students, Khanmigo also offers differentiation, lesson plans, quiz questions, student groupings, hooks, rubrics, and more for teachers. The cost is a modest $4 monthly.

StudyFetch
StudyFetch allows students to upload a wide variety of materials—class notes, PowerPoints, PDFs, study sets from Quizlets, video, audio, and more—from which it generates flashcards, quizzes, games and other study aids, while its AI-powered tutor, “Spark E.” guides students through difficult concepts and encourages deep thinking. Teachers can use StudyFetch to help with lesson planning, presentations, quizzes and even train AI tutor models to their specifications. Free basic account, with weekly ($3.92/week), monthly ($7.99/month) and premium plans ($11.99 to $19.99/month).

Tutor AI
Tutor AI allows students to study nearly any topic based on learning goals, current knowledge, and preferred learning style. Unlike many other tools profiled here, Tutor AI also solves homework problems for students, providing step-by-step solutions with detailed explanations—a practice which may not support genuine learning. For educators, the platform offers the ability to create courses, customized, personalized lesson plans, flashcards and more. Free basic account and two premium levels ($13.33/month, $17.50/month).

Flint
Designed specifically for education, Flint supports student learning by answering questions, offering feedback, and providing reminders for deadlines, time limits, and follow-up activities. Teachers can create or upload content using items such as worksheets, rubrics, lessons, or external resources as needed. Best of all, teachers can set the capabilities of AI actions—from guiding to hinting to explaining—thereby ensuring real learning, rather than AI-assisted parroting. Flint’s free plan allows up to 80 users, while premium plans vary based on school size.

MagicSchool AI
Billing itself as “safe secure AI for schools,” MagicSchool AI’S SOC 2-certified and FERPA/COPPA-compliant platform is aligned with district-level privacy standards and aims to promote critical thinking while teaching kids responsible AI use. Among the more than 50 AI-powered tools for students are those created for feedback, scaffolded learning, and specific task support. Most of the customizable tools seem designed to support genuine learning. For example, the “5 questions tool” generates 5 questions on any topic to deepen student thinking, while “Math review” generates review questions to improve math skills and the AI Learning Assistant provides guidance on multiple topics without giving direct answers. The robust free basic account includes more than 50 tools, quizzes, feedback, and more.

Amira Learning Tutoring and Reading Assistant
Research-based and validated Amira Learning offers a personalized Claude AI-powered reading tutor that allows the unusual ability to listen to students reading aloud, recognize problems with fluency and accuracy, and provide real-time suggestions adapted to the student’s struggles. The standards-aligned, FERPA- and COPPA-compliant platform is available on a district-wide basis.

Class Companion
Class Companion provides personalized feedback of student work in nine subject areas, from English through biology, and extends substantial educator control options, a feature cited by Erik Ofgang as being one of its strongest. He also praised the platforms’ support for the mastery model of education, which allows students multiple opportunities to demonstrate their expertise. Free for educators; however, some features such as LMS integration are available only with a school or district subscription plan.

ChatGPT Study Mode
Rather than simply providing answers, ChatGPT’s Study mode acts as a tutor, offering responses that guide students’ learning without spoon feeding them. Aimed at college students, it employs techniques such as the Socratic method, interactive prompts, scaffolded suggestions, and knowledge checks in the form of quizzes and open-ended questions. When tested by Erik Ofgang, it earned a strong, albeit qualified, “thumbs-up.” Free for all logged-in users on the ChatGPT Free Plan, as well as paid tiers like Plus, Pro, and Team.

Microsoft Reading Coach
Reading Coach provides a personalized reading experience for learners by generating a custom story based on the student’s selection of character and setting. As the learner reads aloud, Reading Coach evaluates their speech and gives feedback on their speed and accuracy. Gamified features such as rewards, badges and streaks contribute to kids’ engagement. Fully free with Microsoft account sign in.

Change is afoot. Or at least that’s the impression I get from the college writing classes I teach, as well as my conversations with educators who teach at various grade levels, and just from generally observing societal trends around education.

As we emerged from the pandemic era in education, we jumped right into the early AI era. Yet as AI skepticism mounts in and outside of education, I think/hope we’re about to enter a new era.

Though readers be warned, when it comes to predictions, I’m not exactly the Oracle of Delphi. After all, I truly thought this year was going to be a good one for the New York Mets. Even so, I do believe all of these predictions have more of a chance of coming true than the Mets do of playing in the World Series anytime soon.

Here’s how I see things playing out in the coming school year.

1. The Techlash Continues

Okay, so I’m no exactly going out on a limb with this one. But I don’t see the ongoing pushback against certain forms of technology going away any time soon.

For example, I haven’t spoken with one teacher from a district or school that banned phones who wants them back--not one. Presumably some exist, but their numbers are evidently small.

Likewise, I believe there’s a sizable consensus that even though screens of various kinds can be helpful, these don’t hold all the answers to education. The emphasis going forward, as it should be, will be on eliminating the tech that distracts from education and focusing on the tools that really and truly help students.

2. AI Cheating Declines

This is a bold prediction, I know. But I see 2026 as the peak of the “I’ll just let AI do my assignment” era that has altered education.

Going forward, particularly at the college level, I think we’re going to see more institutional responses and more cultural responses to this. Encouraging trends include the movement toward watermarking from some of the tech giants, including Anthropic, which recently announced Claude will start watermarking all text it generates.

I’m also seeing a growing sense that passing off AI as your own writing is not a thing that you should do in any situation.

I’m confident the combination of better AI detection models and watermarks paired with more intense societal disapproval will greatly decrease the prevalence of AI cheating by the end of the school year.

Again, this is an optimistic take, but I really think the current AI writing situation can’t go on, and this is the year it finally peaks.

3. Job Anxiety Decreases

The existential dread most of us feel about the future of human-powered labor will subside by the end of this school year, and both educators and their students will breath a collective sigh of relief.

I think this will happen for the same reasons I think AI cheating will decrease. Society by and large has made their feelings clear on what we think of AI-produced work of any kind: We dislike it about as much as we disliked new Coke.

AI will continue to play a key role in research assistance and tutoring, and in areas such as vibe coding. However, as we drive unwanted AI from the classroom, we’ll also drive unwanted AI from professions it has no business in, such as writing, illustrating, and pretty much all artistic fields. This will help students better control their career paths and get rid of that dark cloud on the horizon for some.

4. The Tactile Revolution Will Continue (And Obviously Won’t Be Televised)

The pen and papers that many educators have turned to in a desperate effort to guard against AI misuse are not going back in the desk. Students won’t need to write by hand to prove they’re not using AI going forward, but I don’t anticipate this trend going backward as many educators have realized there is something important and beneficial in the tactile experience. Education in the post-COVID era has gotten a little bit away from it, but more hands-on lessons with objects students can touch and feel has staying power.

For me, I’ve realized the value of getting writing students to create a piece of work that exists in the real, non-digital world. Whether that’s a handwritten note or a printout, changing the medium helps students engage differently, and sometimes connect with their writing in a more meaningful way.

5. The Next Big Thing Is Coming

Forgive me for the vagueness of this one, but with AI use and abuse hopefully maturing in the coming school year and the post-COVID period moving more firmly behind us, I think the field of education is poised to enter a new era of some type. I don’t know what this will be. I just get a strong sense of transition in higher education due to changing demographics and declining enrollment. I think some of that will trickle down to K-12.

I also think this shift will involve some type of truce between the techlash and AI, and hopefully incorporate the best of both worlds.

Ultimately, I’m not sure exactly what the future holds, but I’m confident a new era is dawning. Here’s hoping we can learn some lessons from the past when it does.

Atlas is an AI-powered learning platform designed to help students study more effectively by providing personalized support across subjects. Unlike a simple chatbot, it aims to become a study companion that can answer questions and more. It helps to explain concepts, summarize documents, create revision materials, and even work with a student's own course content.

To minimize the risks of overreliance or academic dishonesty, Atlas positions itself as a platform built specifically for learning, with features designed to encourage understanding instead of simply producing finished answers.

Students can upload class notes, PDFs, presentations, and assignments, allowing Atlas to provide responses based on their own learning materials rather than generic internet information. The platform also supports lecture recordings, helping students revisit lessons and generate summaries or revision resources after class.

For educators, Atlas is an example of how AI is evolving beyond general-purpose chatbots into subject-focused learning assistants. While it shouldn't replace classroom teaching or critical thinking, it can help students review difficult topics, organize information, and prepare for assessments when used appropriately.

Here's everything you need to know about Atlas and how teachers can use it to support learning.

What is Atlas?

Atlas is an AI study assistant designed to support students throughout the learning process. It works across a broad range of academic disciplines, including mathematics, science, history, literature, languages, economics, and more.

Students interact with Atlas through a conversational interface similar to other AI chatbots. However, the platform places greater emphasis on learning by allowing users to upload their own educational materials and receive answers based on those resources.

This means a student can ask questions about a textbook chapter, a teacher's presentation, research papers, or their own lecture notes. Atlas can then explain concepts, summarize information, identify key themes, generate quizzes, or create revision flashcards tailored to that material.

The platform also supports image uploads, allowing students to ask questions about diagrams, handwritten notes, worksheets, or mathematical problems.

Responses can reference a student's own course content, so Atlas can often provide more relevant support than general AI assistants that rely solely on their existing training data.

How does Atlas work?

Atlas uses large language models alongside retrieval techniques that allow it to search through documents and resources uploaded by the student.

After creating an account, you can upload PDFs, Word documents, PowerPoint presentations, research articles, lecture slides, handwritten notes, or images. These materials become part of the student's personal knowledge library.

Instead of answering every question using general AI knowledge alone, Atlas can search those uploaded resources and incorporate these into its responses. This helps students receive explanations that align more closely with the material they are actually studying.

Students can also upload lecture recordings. Atlas generates transcripts, produces summaries, highlights key concepts, and creates revision notes from the recording, reducing the amount of manual note-taking required after class.

Another useful feature is its ability to generate study resources automatically. After analyzing uploaded content, Atlas can produce flashcards, quizzes, practice questions, and condensed revision guides that students can use before exams.

For mathematics and science subjects, Atlas is capable of breaking problems into multiple steps rather than simply displaying an answer. Teachers may find this particularly useful when encouraging students to understand problem-solving processes instead of memorizing solutions.

The platform also supports essay planning and writing assistance. Students can brainstorm ideas, organize arguments, identify supporting evidence, improve structure, and refine drafts.

Atlas can also connect with learning management systems such as Canvas, allowing students to organize coursework and access AI support within the context of their classes.

What are the best Atlas features?

Personalized learning using uploaded resources
One of Atlas's top features is its ability to answer questions using a student's own learning materials. It doesn't just offer generic explanations, but instead can reference lecture slides, course readings, textbooks, and notes that have been uploaded to the platform. This makes responses more relevant to individual classes and assignments.

Lecture summaries
Students can upload recorded lectures and generate summaries, transcripts, key takeaways, and revision notes.

This can help learners revisit lessons, and is especially useful if English is an additional language or if they need extra time to process complex material.

Automatic flashcards and quizzes
Atlas can convert notes and documents into revision materials without requiring students to build flashcards manually. Teachers could encourage students to use these self-generated quizzes as formative assessment before classroom tests.

Step-by-step problem solving
Rather than simply producing answers, Atlas can explain the reasoning behind mathematical calculations and scientific concepts. This makes it more useful as a tutoring tool when students need additional support outside lesson time.

Essay support
Students can brainstorm ideas, organize essays, improve structure, summarize research, and refine written work.

Teachers should make clear expectations around academic integrity so AI supports learning rather than replacing independent writing.

Course organization
Because Atlas stores uploaded learning materials in one place, students can build a searchable knowledge base throughout the school year. Instead of searching through multiple folders, notes, and documents, they can ask questions directly about their existing resources.

How much does Atlas cost?

Atlas offers a free plan that includes daily AI usage limits, allowing students to upload course materials, generate summaries, create flashcards and quizzes, and chat with their documents.

Those who need more extensive access can upgrade to Atlas Pro, which removes many of the usage limits and provides significantly more AI interactions for intensive studying.

Atlas also offers weekly, monthly, and annual subscription options, with prices varying by region and platform. At the time of writing, U.S. subscriptions range from $9.99 per week to $19.99 per month, while a yearly plan is available for $129.99 through Apple's App Store.

Educators should check the Atlas website or app for the latest pricing before recommending the service, as subscriptions and regional pricing may change.

Atlas Tips

• Encourage students to upload class resources
  Students will generally receive more useful responses if Atlas can reference the same materials used in class rather than relying entirely on general knowledge.
• Use AI-generated quizzes for revision
  Ask students to create practice quizzes from lesson notes before a test. This encourages retrieval practice while reducing preparation time.
• Compare AI explanations with classroom teaching
  Have students compare Atlas's explanations with textbook content or classroom instruction. This helps develop critical thinking while reinforcing subject knowledge.
• Teach responsible AI use
  Discuss when AI can support learning and when independent work is expected. Clear classroom expectations help students use the technology ethically.
• Remind students to fact-check
  Like all generative AI tools, Atlas can occasionally provide inaccurate or misleading information. Encourage students to verify important facts using trusted academic sources before relying on AI-generated responses.

AI teaching tools have moved well beyond simple chatbots, with the best options now offer teachers practical help. That can mean lesson planning, assessment, differentiation, feedback, and classroom management. Some put AI directly into students' hands as a learning assistant.

The challenge is knowing which ones are actually useful. Countless AI tools now promise to transform education, but vary wildly in what is offered, designs for schools, and how much work teachers still need to do themselves.

These are some of the best AI teaching tools we've covered at Tech & Learning, selected for the different ways they can support educators and students. Some are designed primarily to save teachers time, while others are built to give students more personalized or interactive learning experiences.

From creating an entire lesson to helping a student practice a speech, these tools show just how broad AI's role in education has become.

Best AI teaching tools

Best AI teaching tool overall

EdCafe AI

Edcafe AI is one of the more complete AI platforms for teachers because it aims to cover the entire teaching cycle rather than focusing on one particular task.

Teachers can use it to plan lessons, create learning materials, assign activities, and assess student progress, all from the same platform. Existing resources can also be uploaded so the AI can work from the material a teacher is already using rather than starting from scratch.

This is helpful when time is of the essence. A teacher can upload a document, presentation, or curriculum material and use AI to turn that into lessons, quizzes, or activities that can then be shared with students.

Tools are also available for differentiation, including chatbots that can be tailored to different abilities or subjects. Teachers can then monitor engagement and performance, creating a feedback loop that can help inform future lessons.

For teachers looking for one AI platform rather than a collection of separate tools, Edcafe AI is a great option.

Best for personalized learning

Flint

Flint takes a different approach by combining an AI teaching assistant with an AI tutor for students.

Teachers can use the platform to create lessons, activities, and feedback, while its Sparky AI tutor gives students individual support as they work. The teacher remains in control of how much help the AI is allowed to provide, which is useful when you want AI to guide students without simply handing them answers.

Teachers can upload worksheets, rubrics, lessons, and other resources and use those as the foundation for AI-generated activities. Students can then work through those activities, ask questions, and receive support along the way.

The personalized element is great here. It doesn't expect an entire class to work through the same material at the same pace, but can provide additional support when individual students need it.

Teachers also get analytics and engagement information, while school-level controls and integrations make this suitable for wider deployment.

If you want to use AI to move toward more personalized instruction, Flint is a top choice.

Best for lesson planning

Eduaide.Ai

Eduaide.Ai is built specifically around the jobs that take up teachers' time, making it a useful AI assistant for planning and creating classroom resources.

The platform can generate more than 100 types of teaching resources, including lesson plans, worksheets, quizzes, assessments, vocabulary lists, problem sets, escape rooms, and role-playing scenarios.

This doesn't rely on an open-ended chatbot prompt, so teachers can select the type of resource they want and provide the information needed to create it. That makes the process more straightforward for educators who don't want to spend time learning how to write elaborate AI prompts.

It also offers a built-in editor, so teachers can refine the generated material without leaving the platform. Support for more than 15 languages adds another useful layer for language teaching and multilingual classrooms.

Eduaide.Ai is less about giving students an AI chatbot and more about giving teachers an AI-powered resource creation assistant. It's ideal when the goal is to spend less time making teaching materials and more time using them.

Best for student AI interactions

SchoolAI

SchoolAI is designed specifically for education, with AI tools intended to support both teachers and students.

Teachers can use it to create tailored lesson plans, quizzes, and worksheets, with automatic alignment to educational standards. Students can then interact with AI in a more controlled environment rather than simply being handed access to a general-purpose chatbot.

SchoolAI is designed around the idea that AI can provide personalized support while still sitting inside a teacher-controlled classroom environment. The platform can adapt learning experiences to individual students, helping provide something closer to one-to-one support without requiring a teacher to spend one-to-one time with every child.

SchoolAI also highlights its FERPA, COPPA, and SOC 2 compliance, giving schools additional reassurance around its use in an educational environment. For teachers who want students to use AI rather than simply have AI create resources behind the scenes, SchoolAI is one to investigate.

Best for creating teaching materials

Almanack

Almanack is an AI assistant designed specifically for educators, with a focus on turning prompts into classroom-ready resources. Teachers can create slide decks, worksheets, lesson plans, quizzes, tests, games, and report card comments. It can also help with longer-term curriculum planning, giving it uses beyond simply creating a single lesson.

One of the top features is differentiation. Teachers can create multiple versions of a resource for different ability levels without having to start again each time. Almanack can also translate resources, which could be useful when working with multilingual students or creating materials for English-language learners.

The platform is standards-aligned and designed so that AI-generated content is created for teachers rather than being delivered directly to students. That leaves the educator in control of checking and refining the material before it reaches the classroom.

Like any generative AI tool, Almanack still needs a human check. But if creating teaching materials is eating into your preparation time, it can take a lot of the initial work off your plate.

Best for standards-aligned teaching

Teachally

Teachally uses AI to create personalized teaching resources while keeping curriculum and standards at the center. The platform can generate lesson plans, teaching guides, and assessments based on the subject, class, and individual learning requirements. It is designed to support differentiation, so teachers can create material that better suits students working at different levels.

One of its biggest selling points is its extensive standards alignment. Teachally supports hundreds of thousands of state and national standards, allowing teachers to connect generated resources with the requirements they're actually teaching. The platform also supports translation into more than 100 languages, giving it potential uses in multilingual classrooms.

That combination of AI generation and standards alignment makes Teachally useful for teachers who don't just want an AI-generated lesson, but want something that fits into the curriculum they already need to deliver. It's another good example of AI being used as a teaching assistant rather than a replacement for the teacher.

Best for public speaking

Yoodli

Yoodli takes AI into a more specific area of education, helping students become better speakers. The platform acts as an AI speech coach, allowing students to practice presentations, speeches, interviews, and other speaking activities privately before performing in front of a real audience.

Students can record themselves speaking and receive immediate feedback on things such as pacing, filler words, word choice, and body language. They can see specific areas to work on rather than simply being told that a presentation needs improvement.

Teachers can also create custom role-playing scenarios and learning activities, while Yoodli's AI Tutor can hold conversations with students and ask follow-up questions. This makes it useful for English, public speaking, presentations, interview preparation, and English-language learners.

The biggest benefit is the opportunity to practice without the pressure of an audience. Students can make mistakes, try again, and improve before they have to stand up in front of the class.

Best for teaching how AI works

Teachable Machine

Teachable Machine is different from most of the tools on this list because the AI isn't simply there to help the teacher. Instead, it lets students actually experiment with machine learning.

The Google tool allows users to train simple AI models using images, sounds, and movements. Students can provide examples, train the system, and then see how it responds to new information. This offers a super accessible way to demonstrate what machine learning actually means. Rather than explaining algorithms purely through theory, teachers can let students experience the process of training an AI system for themselves.

Educational resources are also available that can be used alongside the platform to explore topics such as algorithms, data, and AI ethics. For example, students can create an algorithm designed to identify their preferred type of sandwich, then examine why the system produces particular results. This opens the door to discussions around training data, bias, and the way algorithms can influence outcomes.

And there's no subscription to worry about as Teachable Machine is completely free to use and doesn't require users to sign in. For teaching AI literacy rather than simply using AI, it's hard to beat.

As a longtime educator, I’ve spent a great deal of time designing assessments that promote student choice and voice and enable students to truly show what they know and can do. However, with the rise of artificial intelligence in education, deciding how to assess students has become more challenging.

Students may have access to AI both in and outside of our classrooms. It can generate summaries, write essays, provide explanations, create outlines, and offer feedback within seconds for them. AI can help learning, but it requires educators to think beyond traditional assessments that worked well before AI.

A major concern is that students will cheat or that assessment is broken. Rather than focusing only on how to stop students from using AI, we need to design tasks that make students' thinking more visible. When an assignment asks students to explain their decisions, reflect on the process, make connections, analyze feedback, defend their thinking, and revise their work, learning becomes harder to outsource.

Making this shift matters because if students are only assigned to choose an answer, write a paragraph, create a presentation, or design a final product, AI can do these tasks for them.

Providing guidance for students on how to responsibly use AI is also critical.

Moving Beyond Product-Only Assessment

For years, assessments in my classroom were often built around a final submission such as an essay, project, test, or presentation. I would evaluate it, provide feedback, and move on. These all have value because students build skills in communication, organization, critical thinking, and reflection.

However, what students submit may not fully demonstrate their learning. A well-written essay does not necessarily show understanding. A correct answer does not reveal whether the student can explain the reasoning or simply guessed or used AI. A well-designed presentation does not prove that the student made the design choices on their own. For these reasons, the assessment design needs to be more intentional.

Educators need to look at what the student created, consider how they got to that point, what choices they made in the process, what information was used, how they revised their work, and what I think is important, is what can the student share about their learning. Did the student use AI (if permitted), and how?

Using these reference points will shift assessments from being a simple, quick check of understanding or task completion to showing deeper learning and content retention.

Shifting to More Visible Thinking

Finding ways to make thinking visible takes a little time at first, but it is how to truly see a student’s learning journey. To start, educators can set up a series of checkpoints. Rather than asking students to only submit a final essay, for example, teachers might ask for various artifacts from their work. Students could provide their initial question, a claim, a draft, a peer feedback reflection, revisions made, and a final explanation of what changed. These components encourage students to notice their own thinking.

Educators can also ask students to explain their thinking and choices made. Why did they select this source? Why did they organize the project in this way? Why did they reject one idea and keep another? Why did they make the revisions they did?

These explanations can provide greater insight than the final product alone.

In their work, students can show their thinking through annotations, peer discussions, drafts, concept maps, note-taking, reflections, and even demonstrations. The format may vary depending on the task, but the goal is to have students explain their learning.

Rethinking AI Use and Academic Integrity

A valid concern is that AI makes it harder to understand what students actually know. However, if we ban the use of AI entirely, it won't prepare students to use these tools responsibly.

Instead, educators should create clear expectations around when AI is and is not permitted and model this for students to show how it can be an impactful part of their learning experience. Not every task should involve AI. Students still need opportunities to think, write, create, and problem-solve without relying on a tool. However, students also need guidance in how to use AI thoughtfully, ethically, and transparently.

I have found it to be helpful to design assignments with different levels of AI use. Teachers can build AI expectations directly into assignments. Directions might say, “AI may be used for brainstorming but not for drafting,” or “You may use AI feedback, but you must submit your original paragraph, the feedback, and an explanation of your revisions.” This turns AI use into part of the learning process rather than a hidden behavior.

Establishing a structure such as this helps students understand that responsible AI use is not “all or nothing” but instead depends on the purpose of the task they are being assigned.

Transparency is key. If students use AI, they should be able to say how they used it. Did they ask for topic ideas? Did they use it to clarify vocabulary? Did they compare explanations? Did they request feedback on the organization of their writing? Did they ask for a revision?

Requiring a short AI use note can help students reflect on their choices and help teachers better understand the learning process.

Assessments Should Include Conversations

A powerful assessment tool is conversation. When students talk about their work, teachers can sense understanding, confusion, confidence, and growth in ways that are not always visible on paper.

Not every assignment needs a formal presentation. Having short conferences with students can be meaningful. When a teacher asks a student to talk about the strongest part of their work, an area where they struggled, or even what they would change if they had more time, these questions provide more insight into student thinking. And these are also great ways to build relationships that support student growth.

Students could also record short audio or video reflections explaining their process. I’ve had students do this using Padlet, and it was great for building many skills. Students can engage in peer discussions, small-group activities such as debates, or do gallery walks. These moments create opportunities for students to demonstrate more ownership in learning.

When students know they will need to explain their thinking, the assignment changes. The focus moves from simply producing something to understanding it well enough to discuss it.

Designing Assignments AI Cannot Easily Replace

No assignment is completely “AI-proof,” and that should not be the goal. A better goal is to design learning experiences that are meaningful, contextual, personal, and connected to students’ thinking.

When assignments include local context, classroom experiences, personal reflection, student choice, and live discussion, these become more meaningful and more likely to promote content retention. For example, instead of asking students to write an essay on a historical event, ask them to connect the event to a class discussion, analyze a specific primary source used in class, compare interpretations, and explain how their thinking changed. Or instead of asking for a general science report, ask students to collect observations, analyze class data, explain unexpected results, and reflect on the process.

The more connected the work is to the classroom learning experience and real-world events, the harder it is for students to lean on AI without truly engaging in the learning experience.

It should not be about making assignments more difficult just to “catch” students. By making assignments better so that students have to think, create, question, explain, and reflect, they engage in the process itself and focus less on one final product.

The Importance of Reflection

Reflection may be one of the most important assessment practices in an AI-enabled world. When students reflect, they make their thinking visible. They explain what they tried, what worked, what did not, what they learned, and what they would do differently. They see the process of learning more.

Reflection also helps students develop metacognition. They begin to understand themselves as learners. They notice when they relied too heavily on a tool. They recognize when AI helped them think differently. They learn to evaluate whether a suggestion improved their work or weakened it.

Simple reflection prompts can make a big difference:

• What part of this work best shows your thinking?
• Where did you struggle?
• What feedback helped you improve?
• How did you use AI or other tools responsibly?
• What would you change next time?

The Most Important Outcome of All

Assessment after AI is not about fear, but rather about intentional and purposeful design.

Educators do not need to forget everything they believe about assessment. Many strong instructional practices are already in place: formative assessment, conferencing, portfolios, project-based learning, performance tasks, peer feedback, revision, self-assessment, and reflection. Now, with the availability of AI, these practices are even more important.

The future of assessment will require us to find a better balance between a final learning product and the learning process. Students should still have opportunities to create what they have always created, but they should be able to explain their thinking behind it. They should know when AI is appropriate, when it is not, and how to use it in ways that support rather than replace learning.

For students to be well-prepared for the future, we have to assess how they think as well as what they produce. To do this means designing assignments and assessments that push students to question, decide, explain, revise, reflect, and take ownership of their learning.

AI can generate answers. But students still need to develop judgment.

And that may be the most important outcome of all.

CoChat is an AI-powered tool that provides educators, students, and researchers with a host of capabilities when it comes to conducting research.

While there are other AI research tools, a common problem is citation hallucination. CoChat is designed to cut back on those hallucinations, and that feature is definitely a plus in my experience with CoChat.

Additionally, CoChat is great for deep research. When I searched topics, I was well-versed in, it brought up papers that had taken me hours and, in some cases, days, to track down on my own the old-fashioned way.

After using it for a few weeks, I am impressed with CoChat overall and find it to be the strongest AI deep research tool that I've tried. Here's a closer look at how CoChat works, what it costs, and its best features.

What is CoChat?

CoChat is an AI research tool that harnesses advanced LLMs, including GPT, Claude, and Google, to search more than 200 million academic papers. It offers a feature that allows you to compare different answers from different models and then choose the best one.

In practice, CoChat functions as a highly skilled, very quick research assistant that can find and summarize relevant research on a given topic. CoChat also automatically verifies citations to reduce the frequent AI problem of citation hallucination.

After using CoChat to search several different research areas, ranging from quantum physics to education and health, I have been impressed with its ability to quickly collate research that would take me hours to track down with a tool such as Google Scholar. As with any research tool, it didn’t find all relevant research, though it did a better job than other similar tools I’ve used previously.

More broadly, I still wonder whether there are benefits to the old-fashioned Google-style search. Certainly, the time I’ve spent searching for relevant research hasn’t all been a waste because when I’m doing it, part of my learning process is the search itself.

That said, I do like using a CoChat-like tool toward the end of the process, just to see if this different method turns up anything new.

What Are CoChat's Best Features?

CoChat’s best feature is its robust ability to search through academic research and summarize its findings, all while providing direct links to the original papers. It accomplishes this through access to cutting-edge AI models and, in my experience, seems capable of both speeding up new research and helping expand research I’d previously conducted.

Additional strong features are that it urges you to double-check its work and that its default writing mode is a conversational summary and literature review style. This review summary is helpful, and as an added bonus for educators, is not something a student might be tempted to hand in as their finished paper.

That said, CoChat does frequently offer to help students do more with their research. For instance, after telling me it made a literature review table, it offered to “turn this into a shorter plain-English brief or a tighter source matrix for class use.” When I took it up on this offer, I suspected it would turn the work into a paper that a student could then submit. By doing this, I worried it would move from what I see as an appropriate AI use, research assistance, to an inappropriate one, doing the writing and thinking for a student. Ultimately, those fears were unfounded. CoChat just condensed its findings and used a more conversational tone to explain.

Despite passing this test, when I explicitly asked CoChat to turn its findings into a paper for an assignment "in my class," it was willing to do that, which isn’t ideal.

Even so, given that the availability of AI tools willing to write your papers, emails, and texts is so widespread, it was refreshing that CoChat required this additional step.

What Does CoChat Cost?

The basic version of CoChat is completely free and is what I used for this article. I didn’t come up against any limitations, but there are some metered elements to the free tier such as three file uploads per day and a limit of 10 citation verifications.

Subscriptions start at $9 per month, and there is a pro tier at $19 per month.

The Best CoChat Tips and Tricks

Verify Research

One of CoChat’s main selling points is its ability to check sources it has cited to make sure none are fabricated. Users should take advantage of this feature.

Create an Instant Literature Review

CoChat is as good as any tool I’ve encountered at creating fast and helpful literature overviews on diverse topics.

Use It In Addition To Other Research

Even though it’s tempting to use CoChat at the start of your research, I like using CoChat after I’ve done some research to make sure I found all the relevant material and/or see if it turns up anything that I missed in my old-fashioned search; both happen on occasion. Therefore, using CoChat in this manner seems to deepen my research process overall.

As the academic year approaches, the integration of artificial intelligence into educational settings has shifted from a novelty or optional consideration to more of a necessity.

This summer as I engaged with my students, I had them each evaluate 10 to 12 generative AI tools of their choice. This summary is based on analysis of their notes from recent explorations and highlights a core group of "top-tier" tools that provide substantial value for both instructional efficiency and student success.

By focusing on tools that prioritize research synthesis, administrative logic, and visual communication, teachers and students can navigate the upcoming fall semester with unprecedented productivity.

The following six tools emerged as the most effective according to educators, offering a roadmap for fall implementation.

Gemini Notebook (The Research and Synthesis Powerhouse)

Gemini Notebook stands out for its "grounded" approach to AI, processing only the specific documents a user uploads. This makes it a premier tool for maintaining academic integrity while managing heavy reading loads. Personally, I find this one of the best tools available for summarizing large sets of articles or student papers to find themes or trends.

• How Teachers Can Use It: Educators can upload their entire curriculum unit to freshen it up or mold the materials differently. Student A noted that it can generate podcasts based on uploaded material, suggesting that "instead of having kids read a chapter of a short story, it could turn that into a podcast for you." It is highly effective for creating executive summaries and slide decks from dense research articles to maximize lesson prep time.
• How Students Can Use It: Students can upload lecture notes, textbook chapters, or articles to generate study guides, practice quizzes, or thematic literature overviews. Student B remarked that the audio overview feature is a "game changer," particularly for organizing complex concepts for long-term projects such as a dissertation.
• Concerns: A primary concern is the ethical implication of feeding the beast by uploading copyrighted works into an AI model. Additionally, users must be wary of surface level analysis, as the tool may occasionally overlook important nuances in the source material.

MagicSchool AI (The All-in-One Educator Assistant)

Historically one of my primary recommendations to K-12 educators–it consistently rates a "10/10" by users–MagicSchool AI is a platform designed specifically for the unique needs of the K-12 and university environment. It has a lot of built-in templates and is fully FERPA compliant.

• How Teachers Can Use It: Teachers can utilize a suite of more than 60 tools to generate IEP drafts, PLC agendas, and standards-aligned worksheets. Student C called it a "homerun tool" because it "supports the productivity and direction of [Collaborative Team] time without intruding on the actual meeting."
• How Students Can Use It: The Student AI Tutor offers personalized, 1-on-1 support. Student D found the tutor to be “excellent" because of the "depth of conversation it was having. It told me when I was correct [and] when I wasn’t giving it the skill it asked me to show."
• Concerns: The automated nature of the tool can lead to a loss of human tone if not reviewed carefully. Student E warned that if students rely too heavily on the AI to compile information, they risk "short-circuiting" the struggle required for true critical thinking.

Canva Magic Studio (The Visual Communication Leader)

Canva has evolved from a simple graphic design tool into a comprehensive AI suite that lowers the technical barrier for creating professional-looking learning visuals.

• How Teachers Can Use It: Educators can use Magic Design to instantly prototype course materials, turning a lecture outline into a cohesive slide deck. Student F noted its value for creating "take-home fliers for students to share with parents/guardians" and "quick signs for lessons."
• How Students Can Use It: Students can synthesize complex protocols into clear infographics or collaborative group projects. Student E highlighted that it allows students to "demonstrate competency through multimedia creation, fostering both digital literacy and communicative competence."
• Concerns: A risk of visual homogeneity, in which AI-generated layouts feel repetitive or corporate, exists. Also, there is concern regarding the "Couch Potato" effect, in which students may bypass the critical thinking, research, and synthesis stages if graded purely on the visual artifact.

Consensus (The Academic Search Engine)

Consensus is an AI-powered search tool that helps users quickly identify what peer-reviewed research says about specific questions.

• How Teachers and Students Can Use It: Both groups can benefit from its Consensus Meter, which shows scientific agreement on a topic. Student G found it "amazing" for literature reviews, noting that it "returned sources along with citations, as well as the ability to preview abstracts and PDFs in the same place." Student H appreciated its "excellent" accuracy and ability to find research "without going down a rabbit hole."
• Concerns: A major risk is that students may cite articles based only on the AI summary without reading the full source, which could lead to misunderstandings of the research.

Perplexity (The Citation-Backed Search Engine)

Perplexity serves as an enhanced search engine that provides direct answers supported by current online citations. It has historically been one of my go-to deep search tools, and one of my recent doctoral students used it extensively to gather data for his Delphi Study.

• How Teachers Can Use It: Staff can use it for policy benchmarking or for lesson planning on tangential topics outside their area of expertise.
• How Students Can Use It: Students find it much more trustworthy than basic search engines for report writing. Student I highlighted that the use of citations makes it "highly appealing and ... more trustworthy and applicable to my studies." Student J described it as a "graduate-level Google" that simplifies research into a digestible format.
• Concerns: Because Perplexity pulls from real-time web sources, its results are subject to the errors and possible ethical challenges present on the open internet.

Claude AI (The Writing and Logic Specialist)

Developed by Anthropic, Claude is recognized for its ability to handle massive amounts of text and its focus on providing safe and reliable responses. My students often refer to it as the “poet of the LLMs,” as they think it provides the best editorial assistance when they are writing.

• How Teachers Can Use It: Claude thrives in administrative nightmare scheduling scenarios. Student L described how it caught hidden scheduling conflicts and provided "realistic compromises" when managing transportation, weather, and facility bookings.
• How Students Can Use It: Claude acts as an effective career coach, helping students refine resumes and cover letters while maintaining their personal voice. Student K noted “it was effective because it gave organized and specific feedback on my writing" compared against specific rubrics.
• Concerns: Claude can occasionally produce a generic voice unless the user provides very detailed input. Privacy is also a major concern; users must be careful not to paste identifiable student data into the chat.

Your Fall Implementation Strategy

For the fall semester, these tools can be viewed as collaborative partners rather than replacements for human judgment. As educators, we need to expand the variety of tools that we expose students to in order to ensure they become and then remain digitally literate.

The "80/20 Rule" suggested by Student E is an ideal guiding principle: let AI handle the first 80% of a task, the drafting and organization, while the human expert provides the final 20% of auditing, empathy, and contextual nuance. Of course, the educator and/or student need to know how to develop a good prompt in order to get the AI tools to deliver a product at 80%. That’s one skill that regardless of the tool needs practice to develop.

Another one of my students argued with a faculty member as to why it was okay to ask a human editor to assist them but not an AI one as long as the student was upfront and disclosed the use of AI. Particularly when it is not considered necessary to disclose the use of a human editor or a visit to the writing center for support.

By integrating these tools thoughtfully, educators can reclaim time for high-value relational work while empowering students to become critical editors and auditors of data in a rapidly changing digital landscape.

Thanks to Gemini Notebook and the Winona State University students of EL 590 for their assistance in creating this article.

Even the most ardent AI in education enthusiast will admit that there are times when it’s a mistake to use AI. I’ve been thinking about this a lot recently.

While I never encouraged my students to use AI in any capacity, there were many uses, such as brainstorming, which I was okay with, in theory. These days, I am less open to this and other uses and more wary of AI in general. I’m not the only one.

We’re currently in the midst of what The New York Times recently referred to as a backlash against technology in education. Even among tech-savvy teachers who are AI enthusiasts, I’m hearing more discussion of when it is appropriate or not to use AI.

Based on my own observations and conversations both with AI enthusiasts and skeptics, here are times AI should definitely be avoided in education.

1. Do Not Use AI For: Assessment

Matt Jubelirer, General Manager of Education Marketing for Microsoft, is a big fan of AI use in education in general, as you might suspect. But he told me recently that there’s a growing realization among educators he speaks with that AI is not always the solution.

When it comes to creating tests and assessments, Jubelirer says AI can be really helpful on the preparation side, whether that’s developing a rubric or devising questions, but there’s a limit.

“We shouldn't be using AI to grade people, because we need human judgment,” Jubelirer says. No matter what you feel about AI in general, it's hard to argue with that.

2. For Brainstorming

Adjunct professor Lynn M. Patarini has started asking students not to use AI for brainstorming. “I would prefer folks to use their brain and not rely on AI,” says Patarini, who teaches at the University of Connecticut, Central Connecticut State University, and CT State Community College: Middlesex.

She adds, “I have my students making lists constantly to try and train their brains to think about aim and purpose.”

A year ago, the idea that AI shouldn’t be used for brainstorming may have evoked some controversy. But these days, many of us have witnessed firsthand how AI brainstorming sessions can lead to generic, uninspired outcomes and encourage even well-intentioned students to write with AI guidance.

I know I have. After noticing that AI's advice was making student writing more generic, I’ve recently begun actively discouraging the concept of AI as a writing partner, both among my students and fellow educators.

3. For Lesson Generation

William Henry, an eighth-grade English teacher at Amity Middle School and a freshman composition instructor at Quinnipiac University, both in Connecticut, is a fan of some uses of AI in the classroom. However, there are many instances when he believes it should not be used. One of these no-go AI uses is for lesson creation.

“It could be helpful with generating ideas or ways to implement an idea, but the teacher should be coming up with the bulk of the lesson,” he says.

Lesson planning is a frequently cited AI use case, but as Henry notes, it works best when it's helping an educator execute their vision, not steering the educational ship all on its own.

4. For Emails

This is one that should go without saying, but unfortunately for many students, and even some of those who work in education, it doesn’t.

I have not received AI-generated emails from educator colleagues, but I know people who believe they have. I have recieved many clearly AI-generated emails from students. This might be the most unsettling use of AI I’ve come across.

It may sound like a cliché, but my favorite part of being a teacher is the one-on-one communication with students. Going line-by-line through their writing and offering praise and suggestions, answering questions via email, and connecting with a real human being trying to learn can, at its best, be a beautiful and deeply rewarding experience for all involved. But responding to a lifeless email that was clearly written by a student who put my suggestion into a prompt leaves me cold.

We need to remind students, and even educators, not to be afraid to share their own writing voice with professors, friends, and colleagues. I know some of us are self-conscious about the way we write, but every student I ever had is a better email writer than AI, no matter how much work their spelling or sentence structure might need.

Glint is designed to speed up that process of creating visual resources. The browser-based platform uses AI to generate classroom-ready visual supports from simple prompts, producing resources that can then be edited, printed, or shared digitally.

It is aimed at teachers, special education professionals, speech and language therapists, and families, with a focus on supporting learners who benefit from visual communication.

Formerly known as NerChat, the platform has expanded beyond AI-powered conversation tools to become a broader visual support system. Today, it offers everything from communication boards and classroom schedules to social narratives and interactive visual resources.

This does not aim to replace teacher expertise. Instead, Glint acts as a starting point. Educators can create a resource in seconds before refining it to suit an individual learner, saving valuable planning time while still allowing for personalization. That makes it very useful in busy classrooms where multiple students could need different levels of support.

What is Glint?

Glint is an AI-powered tool designed to help teachers create visual supports, communication boards, and social stories for students with additional learning needs. Formerly known as NerChat, the platform has evolved into a broader resource creation tool focused on helping educators and support professionals build personalized communication materials faster.

Creating visual resources can be an important part of supporting students with autism, speech and language needs, and other learning differences. However, designing effective schedules, symbol boards, and social narratives can also take significant planning time, especially when teachers need to adapt materials for individual learners.

Glint aims to reduce that workload by using AI to generate editable classroom resources in seconds. Teachers can create materials for specific routines, activities, or communication needs, then adjust any to better suit their students before sharing digitally or printing for classroom use.

With features such as text-to-speech, multilingual support, and QR code sharing, the platform is designed to make visual communication tools more accessible across different learning environments.

How does Glint work?

Glint uses generative AI to build visual learning resources from natural language prompts. A teacher simply describes what they need -- for example, a morning routine, classroom expectations, a communication board for snack time, or a social story about visiting the library -- and the platform generates an editable visual support within seconds.

Glint produces a starting point that educators can adapt. Teachers can edit text, replace or rearrange symbols, adjust layouts, and tailor resources to suit an individual learner's communication level or educational goals. This flexibility helps in SEND settings, in which even students with similar diagnoses may require very different supports.

Resources can include images, symbols, written text, and in some cases, speech output, to encourage communication and independence. Finished materials can be printed as traditional classroom resources or shared digitally, while QR codes make it easy for students, parents, and support staff to access any from a phone or tablet.

Because Glint is entirely browser-based, there is no special software to install. It operates across Windows and Mac computers, Chromebooks and iPads, making it accessible in most school environments. Schools can also manage users centrally, helping larger organizations roll out the platform across multiple classrooms while maintaining consistency between staff members.

The overall experience is designed to minimize preparation time while still giving teachers full control over the finished resource, ensuring materials remain appropriate for each learner rather than relying solely on AI-generated content.

What are the best Glint features?

The best feature of Glint is how quickly it produces resources that would traditionally take far longer to create manually. Visual schedules, communication boards, and social stories can often be generated in less than a minute. This allows teachers to spend more time working directly with students instead of designing classroom materials.

Support for more than 50 languages is another standout feature. This can help schools communicate more effectively with multilingual families while also creating resources that are accessible to students who are learning English or who use another language at home. Being able to generate matching materials in different languages also helps reinforce routines across school and home environments.

Interactive communication boards go beyond static printable resources by incorporating text-to-speech functionality. Students can select symbols to hear words or phrases spoken aloud. This can support communication development and promote greater independence for those who benefit from augmentative and alternative communication approaches.

Glint also offers flexible sharing options. Resources can be exported as PDFs for printing, displayed digitally on classroom devices, or accessed through QR codes, making it straightforward to distribute materials without additional software.

For schools and districts, administrative tools allow accounts to be managed centrally, simplifying deployment across multiple classrooms or campuses. Combined with the ability to save and reuse previously created resources, this makes it easier for teachers to build a growing library of personalized visual supports that can be updated rather than recreated each time.

How much does Glint cost?

Glint includes a free tier that lets you generate up to eight visual supports each month, making it a good way to try the platform before committing to a subscription.

For individual educators, the Monthly plan costs $18 per month (discounted from $25) and can be canceled at any time. The Annual plan costs $150 per year -- equivalent to $12.50 per month -- offering around a 30% saving compared with paying monthly. Both paid plans include premium languages, access to the in-app Google Images library, the Glint mobile app, QR code sharing, and advanced board editing tools.

Schools and districts requiring multiple licenses can request custom pricing. These organization plans add centralized billing, shared resource libraries, administrator controls, usage insights, Clever/SSO rostering, and onboarding and training to support larger deployments.

Best Glint tips and tricks

Start broad
Begin with a simple prompt to generate a draft resource, then personalize it for individual students rather than trying to create the perfect prompt first time.

Support families
Use the multilingual tools to create matching resources for home and school. This can help reinforce routines and improve communication between educators and caregivers.

Build a resource library
Save commonly used schedules, communication boards, and social stories so these can be quickly adapted for different students instead of recreated each time.

For Matt Kuhn, Chief Technology Officer for Volusia County Schools in Florida, artificial intelligence has been around longer than most folks realize.

“The first AI was really Clippy way back in the day, right?” Kuhn says. “In the 90s when Clippy would jump on your screen and kind of annoy you a little bit, but it was gathering feedback from you and then differentiating what it told you next in terms of assisting you in learning how to use the program.”

Like Clippy (but dramatically less annoying), Kuhn has been active in shaping how AI is being used in his district, particularly by students.

“One of the things we've learned over the last couple of years is when we talk about using AI, we talk about it being a coach, not an answer giver, or we also say that it's an augmentation to your learning,” Kuhn says. “We want all of our AI to let students go through that healthy struggle but keep them right on the edge much like a video game does. You know, when a student loves a video game, it's because that video game has mastered that moment where the student feels like with just a little more effort, I could get to the next level. That's what we want AI to also do.”

Kuhn, a former small district teacher and principal before a stint as an author and edtech consultant, was also instrumental in Volusia’s early adoption of AI.

“We had one of the first policies around AI in the state of Florida in a school district, and I know it was one of the first because a lot of districts in Florida and even other states started contacting me and asking for a copy,” he says. “Our policy took a pretty open approach to allowing AI with some guardrails put in place so that people knew what the structure of it was and how far those boundaries could be pushed. But we wanted to make sure that our teachers and our students had the freedom to innovate within certain parameters.”

Kuhn has continued to lead AI efforts in his district, which include the creation of an AI professional development academy as well as the deployment of AI agents for district leaders to help them be more efficient and effective in their work.

For this and other efforts, Kuhn was recently recognized with a Tech & Learning Innovative Leader Award at the Tech & Learning pre-ISTE Leadership Summit in Orlando. He shares his district’s AI efforts for students, teachers, and staff.

AI Is A Human Endeavor

Volusia County Schools is a diverse district of 60,000 students and 8,000 employees, with about 115 in Kuhn’s IT division. And with that many involved, there was a lot of concern about how to approach AI.

“We had to alleviate their fears with training and other data to show them that AI could be a positive experience for their students,” Kuhn says. “There's still certainly people that are skeptical, but our data shows that it's being used more and more each year by our teachers and our students. And we're an A-rated district for the second year in a row by the state of Florida. So our academic scores and other indicators are going up despite the fact that our AI usage is also going up.”

To address concerns and support staff who were interested in using AI, the district created its AI Academy, a year-long program that provided ongoing professional development events rather than a one-time workshop. The academy offered differentiated learning paths for various roles, including clerical, educator, and administrative staff, and ultimately, was a hit with everyone.

“The academy kept getting so big that we kept having to change rooms to larger venues just to accommodate all of the people who wanted to go through it,” says Kuhn, who added that the experience needed to be in person. “We found that face-to-face was going to be best for this, where people had their computer in front of them and we could walk around and assist and get them collaborating. We also wanted people to realize that AI is also a human endeavor, that it's not just about technology.“

Vendor partners were also part of the program. “If something new was coming out in Copilot, for instance, we would arrange with our Microsoft customer support to teach us that latest and greatest thing so that we could roll that out right away,” Kuhn says.

The academy concluded with “Volusia Learns,” a week-long district conference during which participants received digital badges. “And now when you see their digital signature or the directory information on their school website, they're showing the badge they earned,” Kuhn says.

Agents On Board

Similar to many districts, trying to navigate and analyze data from numerous disparate sources was a growing obstacle for Volusia schools. Director of Educational Technology Melissa Carr challenged the IT team to make board policies, procedure manuals, student data, and other info that’s typically housed in different departments and silos easily accessible to all.

“In the old way of doing things you have to kind of ask everybody ‘Where do you find this? Who has the latest on this? Can they email it to me?’ says Kuhn. “She wanted it to be a quick thing where a principal could just type in some things and this stuff would appear.”

To solve that challenge, Kuhn and his team developed the Volusia County Leaders Agent, a custom AI agent using Copilot 365 that could access district and student data, board policies, and procedure manuals, all with specific naming conventions for easy indexing. They then encouraged district leaders to customize it.

“Essentially what we did is we taught them how to make their own agent,” says Kuhn. “We put together a set of instructions and we gave them the links to the SharePoint site and the URLs that we would be using, and we walked them through. It was only three pages with screenshots, so it's not like a really hard process to create an agent.”

Having the leaders make their own agents allowed them to be very specific in how they designed it, and also fostered ownership and buy-in.

“They're like, ‘Oh, I've got this agent I made,’” says Kuhn. “‘Look at this. It does this cool stuff for me.’ You know, so instead of just saying, ‘Here, use this that we gave you,’ they're proud of it."

Keeping ’Em Swimming

The district plans to continue offering PD opportunities around AI, although it may not be another full year of the AI Academy as there is limited time available for training and many implementations happening in any given year.

Nonetheless, Kuhn and his staff are identifying and training Digital Learning Teacher Leaders (DLTLs), who lead deep, ongoing training on the latest AI tools with staff members within their building, and also provide resources and support. This approach helps them try to keep up with the pace of AI’s evolution.

“It evolves so fast,” says Kuhn. “But the way we figure is if we at least get them in the pool and swimming, even if the temperature of the water changes a little bit, they'll be able to deal with it.”

Tools They Use

• Microsoft Entra ID (Azure Active Directory)
• Microsoft 365 (Outlook, Teams, OneDrive, Word, Excel, PowerPoint, OneNote, Forms)
• Microsoft Entra ID (Azure Active Directory)
• ClassLink / VPortal (single sign-on & app catalog)
• Canvas (Learning Management System)
• Focus SIS (student information, grades, attendance, analytics)
• Focus ERP (HR, payroll, finance, purchasing, budget – 19 modules)
• IncidentIQ (help desk, work orders, asset management)
• Jamf (Apple device management)
• Cisco Meraki (network switching & wireless)
• Cisco Identity Services Engine (network access control)
• ContentKeeper (web filtering & student internet safety)
• Finalsite (district & school websites)
• BoardDocs (school board agendas & meetings)
• ParentSquare (family/school communications tied to SIS)
• Canva (instructional and staff content creation)
• IXL
• Achieve3000
• Big Ideas Math
• Khan Academy
• Desmos
• PhET
• BrainPOP
• Facilitron (facility scheduling & community use)
• Lightspeed Web and Classroom Filter

While researching my recently updated article, Best Sites for Blended Learning, I came upon CK-12, a remarkable, new-to-me teaching platform that provides a varied and robust set of digital textbooks, AI-powered digital tools, teacher agents, simulations, and more.

CK-12 is the brainchild of the nonprofit CK-12 foundation, whose mission is to provide free access to high-quality, customizable educational content in multiple modalities, suited to multiple student learning styles, levels, and resources. Even better, all tools and features of CK-12 are fully free to educators, students, and families.

What is CK-12?

CK-12 is a K12 online education platform designed for teachers, schools, and families that serves as a customizable digital platform for teaching and learning.

Its flagship offering, Flexbooks, are standards-aligned digital textbooks for K12 math, science, English, social studies, and other topics, including college-level and adult education. Flexbooks are available in English, Spanish, Korean, German, Chinese, Greek, and Polish. A small selection of Flexbooks include built-in interactives.

In addition to Flexbooks, CK-12 offers a wide range of AI tools for teachers, AI teacher agents, learning analytics, study guides and adaptive practice for students, interactive simulations, and PLIX (Play Learn Interact eXplore) interactives.

How Does CK-12 Work?

Users can easily sign up for CK-12 via their Google, Microsoft, Apple, or email accounts. Once signed in, the teacher’s homepage highlights the newer AI-powered tools highlighted at the top, while the older flagship Flexbooks are lower down.

While the AI-powered tools and agents are intriguing and potentially time-saving for educators, the core of CK-12 is really the digital textbooks known as Flexbooks.

(Note that the continual right-to-left and left-to-right animation of the AI Teacher Tools boxes is an unfortunate choice for an organization that aims to be sensitive to learners of all types. Fortunately the student version of the site doesn’t include such distracting animations.)

FLEXBOOKS

Flexbooks carry a number of advantages over traditional textbooks. First, these are fully free, which offers an opportunity for substantial savings—perhaps as much as $1.5-2 million dollars for a district. And in contrast to paper textbooks, Flexbooks can be updated as needed, which of course represents yet another financial saving.

Additionally, Flexbooks are designed to be compatible with all external screen reading apps and tools, allowing for all content to be read aloud. This is especially useful for students who struggle to read at grade level. You can search Flexbooks by grade, language, and subjects. Additionally, one can browse the list of all Flexbooks, arranged by topic and grade.

For six selected math topics, Flexbooks are also available in an interactive as well as conventional version.

Educators can create classes directly in their CK-12 account, or via an existing LMS class, such as Google Classroom, Canvas, Schoology, and others. Then add students through email, roster upload, or direct entry. Assignments can be made quickly through links on every content page.

AI TEACHER TOOLS AND AGENTS

CK-12’s AI-powered Teaching Assistant comprises a large library of tools to speed teaching preparation, assessments and assignments, specialized learning, communications, and much more. Teacher Agent goes further, allowing users to quickly set up their own AI-powered agent, capable of interactive with students while adapting to their responses and learning styles.

What Are The Best CK-12 Features?

In addition to its flagship product Flexbooks, CK-12 offers several key customizable features ideal for the classroom setting.

ADAPTIVE PRACTICE

Adaptive practice exercises in selected topics are available for students in grades 1-5. Each topic includes dozens of subcategory skills that can be assigned to students.

For example, the Measurement topic is divided into money, time, systems, conversions, capacity, and mass. Within each subcategory skills include comparing, identifying, converting, and many others.

FLEXBOOK TEACHER TOOLS

CK-12’s Teacher Tools are built into every Flexbook page and include a quiz creator, lesson planner, and lesson “hook” generator. Simply click the desired link in the lower left-hand box in order to create a quiz, standards-aligned lesson plan, or lesson hook for the content on the current page.

As with most CK-12 features, users are free to adapt the pre-filled templates according to their own preferences.

LEARNING ANALYTICS

Comprising “foresights” and “insights,” Learning Analytics provide granular data on student and class performance, as well as overall performance for each question. While Foresights predicts a student’s future proficiency in the selected topic, Insights displays skill levels that students have already demonstrated, from "beginning" through “mastery.”

Using graphs and lists, the platform also shares performance data from the class, makes recommendations about how to address performance gaps, and highlights students who are performing exceptionally well.

INTERACTIVE SIMS

Sims are a great way to combine text- and graphic-based learning with interactivity, engaging students more deeply than static lessons. Available for physics and chemistry topics, CK-12’s interactive sims are concept-driven, standards-aligned, and offer downloadable PDF worksheets.

Each assignable simulation begins with a brief animation representing the target concept and includes a video tutorial, challenge questions, and easy sharing options.

How Much Does CK-12 Cost?

Remarkably, CK-12 is completely free for all users -- educators, students, families or anyone wishing to study and learn. The nonprofit CK-12 Foundation’s core mission is to provide content and technology tools that support customizable teaching in multiple modalities, for all types of learners, therefore all its tech tools are 100% free.

I was writing a speech for an end-of-the-year ceremony recognizing my students moving up to their next school. It’s one of many speeches, announcements, recommendation letters, and notifications I compose on any given day.

I entered a quote I had written into an AI prompt with instructions to produce a concise, inspiring speech for my fourth graders moving up based on it. I was impressed with the output: It was well structured, followed my conventions, and did a decent organizational job. Yet, it was missing the most important aspect, something I call “personification.”

The biggest mistake users make using AI to write is overlooking the visible lack of substance and eerie wording in LLM outputs. In fact, according to a study by researchers at four U.S. universities, AI chatbots have linguistic fingerprints that reveal which model is used: In five popular AI chatbots, distinct verbal quirks emerged.

It is the personification a human adds that’s absent in AI outputs. Understanding this can help everyone in your school community write better, not using AI as a crutch but rather as a partner–a partner who can put together good looking, well-structured passages, but with the need for the human touch to personify it.

The Personification Two-Step

Let me illustrate with a common example of what many school leaders are writing: letters of recommendation.

In years past, writing a good letter of recommendation would take me, a quick writer, 25-30 minutes to compose. Not bad, but when you add up the half dozen requests all happening at the same time, along with all my other writing tasks, it gets overwhelming. Normally, writing those six letters alone would take three hours.

Now, I take a copy of the requester’s resume, remove their name and identifying information, and ask the AI to write me a letter of recommendation. I can add into the prompt, positive, praiseworthy tone, or more neutral for someone with whom I am not so impressed.

Using this prompt and an attached resume takes 2-3 minutes to get an output. I will add their name and title that were previously removed. But the biggest gap is the personal touch: personification. This is where I go back and take very important steps.

1. I shift the tone from a rigid style even an AI that claims to know my writing style has. This reinforces that the human element has to be integrated and;
2. I add unique details only I know and on which I can elaborate. For instance, that one time the teacher stayed until 6:30 PM to make sure a student of theirs was picked up when the family had an emergency. I can speak to the empathy and character of the teacher in a way the AI neither knows nor understands.

As far as writing articles like this one, I remain dutifully old school. I prefer to write from inspiration, and then go back and edit. I may add a sentence, possibly even a passage that was adapted from an AI output, nothing more.

Don’t get me wrong–I have tested a concept article with AI. And then I scrapped it. The AI cannot give me the natural, conversational tone that I can. To me, good writing means the author is speaking to you. An AI speaks at you, kind of like an instruction manual.

As far as recognizing my tone, the AI can guess at my intonation, but not in a consistent manner. Here, you can feel it come off the page, like a story. An AI puts words together in a calculated, meaningful way but that substance is persistently missing. This is the nuance that is just plain absent.

Crafting A Partnership

As a school leader and author, I am selfishly happy AI doesn’t get nuanced, because that means learners and leaders like me must continue to develop the very abstract skill of writing. The exercise helps me work out my mental muscles.

Another significant challenge with relying on AI for writing is the odd tendency to mash nearby paragraphs together, often creating inaccurate or just plain false assumptions, or vaguely near estimates. AI outputs can give us help with those early questions when we query, offering options we may not have considered but when we dig deeper, when we really write from nuance, humans clearly own writing. Ignoring this will only illuminate AI overuse in student or adult outputs. AI lacks that embodied self awareness.

An MIT meta-analysis confirms this: the combination of human and AI works best when each partner can do the thing they do better than the other, and there is a synergy in this partnership. AI excels at rapid data processing and drafting (essentially calculations), while humans add that critical thinking, empathy, and final context. Humans must check the output and revise, remove linguistic fingerprints, and personify the content.

Understanding this balance provides a hybrid method that helps us work better without losing our natural thinking skills. Ignoring it creates obvious, almost embarrassing attempts at professionals intent to be positive, such as this letter an administrator emailed, copied straight from a chatbot.

The Bottom Line

I have found AI to be a helpful and efficient writing partner for quick starter thoughts and organizational structure. However, human nuance and embodied self-awareness win every time when it comes to qualitative and intuitive writing. AI can assist with brainstorming, research and editing, but it cannot replace authentic emotions, life experiences, imagination and unique voice.

Consider this as a guideline for reminding students and school communities that writing is one of the most important critical thinking skills to help develop the mind and express creative thought.

While generative AI can automate administrative tasks, streamline lesson planning, and free up time for student interaction, it also introduces new safety risks. Addressing the Risks of AI-enabled CSAM and Explicit Content in Education, Qoria’s global report, shows that these tools are also being exploited to produce child sexual abuse material (CSAM) and explicit imagery with alarming ease.

The weeks before the school year begins offer a critical window for district leaders to move from a reactive posture to a proactive strategy, taking strong digital safety steps before students return for the first day of school.

From the Data: Understanding the AI Digital Threat

School leaders should first understand today’s digital exploitation. Catherine Brown, Chair of the Internet Watch Foundation (IWF), says, “AI image generators have become much more sophisticated, to the point where almost anyone can make or manipulate life-like sexual or nude imagery of children at the click of a button.”

So-called "nudifying" apps have dramatically accelerated the ease with which this material is made, leading to devastating instances of grooming, distress, and bullying. The data gathered in Qoria's global survey underscores that these problems are active on school grounds today:

• Widespread Administrative Concern: An overwhelming majority of school leaders–91.4 percent in the United States, 90.5 percent in the United Kingdom, and 91.7 percent in Australia–voiced deep concerns over adult perpetrators using AI to groom students.
• Monthly On-Campus Incidents: Around a third of respondents in the U.S., the U.K., and Australia reported experiencing at least monthly incidents in which students possess, share, or request explicit, child-generated nude images.
• Younger Average Ages: The average age for these explicit content incidents is currently concentrated between 11 and 13 years old. Even more concerning, approximately 12 percent of U.S. and Australian respondents reported seeing these behaviors in children as young as 8 to 10 years old.

Despite these figures, few schools have preparation and prevention measures in place. More than 30 percent of the surveyed school leaders admitted they are unfamiliar with the common online grooming tactics that modern perpetrators use.

When school personnel do not understand the technical mechanics of the threat, early detection becomes nearly impossible.

Actionable Step 1: Empower Teachers Through Specific Safety Training

Teachers are a school's front line of defense, yet they are frequently left out of high-level technical digital safety briefings. To build an abuse-aware culture before the first bell rings, district leaders should design back-to-school professional development that addresses AI safety directly.

First, training must explicitly cover how AI changes grooming behaviors. Traditional grooming relies heavily on human text interaction over long periods. Modern AI algorithms allow bad actors to analyze massive amounts of social media data to identify vulnerable children with absolute precision. These tools can map out student interests, emotional states, and communication patterns to generate deeply persuasive, personalized messages. Staff need to know that a student's sudden emotional withdrawal or obsessive device usage could be connected to highly sophisticated online interactions.

Second, training must address the internal threat of student-generated content via nudifying apps. Whether a student acts out of thoughtless experimentation or clear intent, their developmental immaturity may blind them to the dual damage of these actions: the significant harm inflicted on the peer depicted and the severe fallout to their own relationships and future reputation. Teachers must be instructed on strict reporting protocols, understanding exactly who to contact in IT, administration, and school counseling the moment they become aware of a simulated image circulating in a peer group. This approachable guide for schools on AI-generated sexually explicit imagery from OCEPI can help with a framework for creating plans for both prevention and response.

Finally, training should acknowledge that staff safety matters, too. The Qoria report notes that there have been instances in which teachers and school staff find themselves victims of image-based manipulation through the misuse of AI technology. Treating this as a developing human resources concern builds trust and reassures educators that the district protects its entire community.

Actionable Step 2: Build Proactive Parent Awareness

The digital lives of students are happening both during the school day and when they leave school property. In fact, the risk can escalate during unsupervised time at home. Qoria's report indicates that a quarter of children are already using AI apps regularly to complete their schoolwork, meaning these tools are a constant presence in their domestic lives.

While parents might be focused on screen time and cyberbullying, AI-generated explicit content might be a blind spot. Districts can use the pre-school orientation window to actively educate parents.

A successful parent awareness campaign before the school year begins should focus on transparency and trauma-informed child-centered practical tools:

• Host Digital Literacy Orientations: Do not wait for a crisis to inform parents. Provide specific back-to-school webinars or easy-to-read guides detailing the reality of generative AI tools and the ways the school is supporting their children. Avoid alarmist language. Instead, explain the mechanics of image generators and peer-to-peer messaging apps clearly. OCEPI has a 10 Best Practices guide with a clear framework for how to structure messaging.
• Provide Parental Control Frameworks: School districts should offer concrete guidance on how families can manage devices at home. Introduce parents to reliable parental control tools and explain how to set up filtering on home networks. ConnectSafely has parent guides with clear instructions that school leaders can share with families.
• Create Clear Communication Pipelines: Parents need to know that if they discover a deeply concerning digital incident at home, the school is a safe partner rather than a punitive authority. Provide families with clear, step-by-step reporting instructions to ensure open dialogue. The Cybertipline from National Center for Missing and Exploited Children (NCMEC) is helpful to share.

Actionable Step 3: Provide Meaningful Layered Student Support

Protecting students requires a combination of clear education and protective technology. Before the school year starts, administrators must audit and align both educational and technical sides of student support.

On the educational side, digital wellness curriculum must expand far beyond plagiarism or screen-time management. When designing lessons regarding AI-generated explicit content, districts should follow the 10 best practices from the Online Child Exploitation Prevention Initiative (OCEPI).

First, curriculum developers must minimize or eliminate fear-based messages. Learning about worst-case outcomes seldom changes youth behavior because children typically believe it will not happen to them. Next, lessons should emphasize the social and emotional consequences of these behaviors. Focusing strictly on legal warnings or telling students that certain digital actions are illegal actually decreases the likelihood of reporting.

Additionally, student programming must maintain a child-centered approach. If a student is targeted by a nudifying app or an online predator, the school messaging must reassure them that they are not responsible for what happened. Shaming or blaming a student who has been manipulated only silences them.

True prevention requires active skill-building. Programs should explicitly teach how to set digital boundaries, utilize assertive communication, take screenshots as a record, and report harmful behavior. Because introducing these sensitive topics frequently prompts real-time disclosures, counselors and support staff should be on hand to receive student reports with immediate empathy and support.

IT departments have a role on the technical side. Before students arrive for the first day of school, technology teams can ensure that filtering and monitoring software are fully updated to flag searches for known nudifying applications and explicit keywords. Early detection is absolutely critical. The Qoria report emphasizes that many technologies schools already possess, including content filters, student monitoring systems, and digital wellness check-in tools, can assist in the early detection of risk.

Stepping Into the Future Together

The challenges introduced by advanced generative AI can feel daunting to already overwhelmed educators and parents. However, school and district leaders have the authority and community trust required to get ahead of these digital threats. Intentional steps during the summer to train staff, engage families, and strengthen network visibility can cultivate an environment where technology serves student learning without compromising student safety.

As the Qoria report concludes, planning for a unified community approach ensures that no individual school leader, teacher, or parent has to navigate the risk of AI-generated exploitation alone.

Resources

• Addressing the Risks of AI-Enabled CSAM and Explicit Content in Education. Quoria. 2025.
• Guidance for School-Based Professionals and School Leaders. Online Child Exploitation Prevention Initiative (OCEPI). 2025.
• Ten Best practices for Prevention Programs. Online Child Exploitation Prevention Initiative (OCEPI). 2024.
• Parent Guides from ConnectSafely. ConnectSafely.
• Cybertipline. National Center for Missing and Exploited Children (NCMEC)

TeacherServer is a teacher-created and -run AI platform that offers thousands of free tools for educators. More significantly, the whole platform complies with Department of Education guidelines, offers the option of adjusting outputs to various grade levels depending on your students' needs, and all information is private and deleted immediately after you use it.

After experimenting with TeacherServer, I found it to be a wonderful resource for various AI tools, from quiz and lesson plan generators to subject-specific resources, including data analysis, spelling and grammar checkers, and much more.

Because it is teacher-backed and education-aligned, I felt better about using TeacherServer than most AI tools. Even so, I still found some similar frustrations to when I have experimented with AI in the past: namely, that despite what I hear from colleagues, I never seem to get great results from AI when it comes to lesson planning or slide generation.

Here’s a closer look at TeacherServer.

What is TeacherServer?

TeacherServer is a free library of AI education tools that are designed for teachers. It was created by Zafer Unal, a professor and associate dean of undergraduate studies at the University of South Florida. Unal, who has a dual background in computer science and education, designed TeacherServer to address concerns educators have about using AI. To that end, TeacherServer is focused on data privacy and grade-appropriate learning. It’s also quite fast at generating materials thanks to its utilization of a locally installed AI server.

TeacherServer comes preloaded with more than 1,194 AI tools. These tools are utilized by other teachers, and some are ranked based on popularity. Tools are separated by subject, and users can often input what grade level they are teaching for more grade-specific responses. While TeacherServer’s initial focus was on K-12 resources, it also offers tools for those in higher education.

A truly dizzying number of capabilities are available, from talk-to-PDF features to all manner of subject-specific quiz, lesson plan, and summary generators. While you could do some of these tasks with a standard AI tool, doing so and getting good results would often require some prompt engineering. The AI tools at TeacherServer start already prompted for education overall, and for specific education tasks in particular.

What Are TeacherServer's Best Features?

The education-focused nature of all the AI tools offered on TeacherServer is what stood out to me. Additionally, the focus on privacy, speed, and education standards is really important.

Another fun feature is that simply browsing the catalog of AI tools can give teachers ideas about how other educators are using AI and possibly inspire them in the process.

What Are Some Areas Where TeacherServer Could Improve?

As mentioned at the start of this article, as much as I loved TeacherServer overall, I did have some of the same complaints with TeacherServer as I do with AI overall.

Despite what I hear about how others are using it to save time, I don’t find that that’s the case for me unless I am willing to really compromise on quality. For instance, I’ve never found AI to produce really good slides. True to that trend, when I asked TeacherServer to summarize a link to a resource story, I thought the slides it created were fairly basic and not very helpful.

Additionally, while a grammar and spelling tool was good at catching grammar mistakes, I still did not like the way AI tried to edit my writing. It tried to get rid of any character in my writing, and more unforgivably, it had some incorrect suggestions on comma use and tried to change content within quoted passages.

How Much Does TeacherServer Cost?

TeacherServer is free. To use it you just need to go to TeacherServer.com and create an account. This accessibility is part of its mission and a great feature for educators.

What Are The Best TeacherServer Tips and Tricks?

Search TeacherServer By Subject

TeacherServer has a wide variety of subject-specific tools that teachers of various topics can explore. These range from music and art lessons to math, writing, and more.

Search Most Popular Tools

Exploring the most popular tools on TeacherServer gives you a quick idea of what is working for other educators. This can save you time and give you ideas for prepping in your classroom.

Be Critical

TeacherServer is a really teacher-friendly AI tool, but that doesn’t mean it's perfect. You’re the human being in the classroom and should still be carefully vetting any outputs from any AI tools, including those from TeacherServer.

Copilot Study and Learn is a mode Copilot can operate in that employs Microsoft’s “study and learn agent,” which is built on learning science practices and designed to encourage active thinking rather than give away answers. I recently tried it and was impressed overall. It does a better job than Copilot in its normal mode at guiding students through the learning process.

Here’s a closer look at how to use Copilot Study and Learn, what I like about it, and where I felt it could still use some work.

What is Copilot Study and Learn

Copilot Study and Learn is an AI agent within Copilot specifically designed for learning. Copilot Study and Learn is programmed to incorporate concepts from learning science, including adaptive scaffolding, productive struggle, active learning, and application and transfer.

The tool is available for all education subscribers to Copilot. Activate it by opening Copilot, clicking the plus sign in the left-hand corner of the prompt box, and selecting the Study and Learn option.

Where Copilot Study and Learn Earns High Marks

Copilot did a wonderful job of refusing to do the work for me. It never turned into an answer machine, which most chatbots do and I have found to be an issue even for some study and learning AI chatbot modes I’ve looked at in the past.

Specifically, Copilot Study and Learn did a nice job giving me a brief lesson on elementary particles. Elementary particles are, as Copilot put it, not made of smaller pieces, “as far as we know.”

Copilot began the lesson by asking me what I knew from what atoms are made. After I said, “I know atoms are made of protons, which consist of quarks and something else, and electrons,” Copilot responded:

Nice—you’re already thinking at the right level 👍

You said atoms have protons (made of quarks) and electrons. Let’s sharpen that a bit:

Protons are made of quarks

There’s also another particle in the nucleus… (you hinted “something else”)

Electrons are not made of smaller pieces (as far as we know)

Similarly, when I entered a writing prompt I use with my Composition 100 college students, it actually provided really good advice on putting together the essay.

By comparison, last time I tried ChatGPT in study mode it would almost always end up writing material for me as "an example." Even though ChatGPT kept telling me it wouldn't write my essay, that's exactly what it did.

Copilot did not do this and, in fact, refused to write the paper for me even when I tried to trick it into writing my paper by asking it to write it as an example.

Where Copilot Study and Learn Could Use Additional Supports

Despite its successes, my experience with Copilot Study and Learn was not all positive. Copilot provides constant links to back up the lessons it gives. This seemed like a major plus at first, but annoyingly, many of these links have to be copied and pasted and are not shared as URL links. Worse, many links--at least 25%--did not work. Copilot also had a tendency to refer to illustrations within these links, which compounded the problem. Of course, using visual aids is a great teaching tool, but not if the links to the illustrations don't work.

More broadly, as impressive as Copilot Study and Learn and other similar tools are, these have an inherent limitation when it comes to motivation. I found the more I used these tools, the more bored I became. Learning is hard, and I’m not saying a new chatbot should solve that problem entirely, but to be effective these tools need to be at least as engaging as existing learning technology, such as YouTube videos or gamified learning apps such as Duolingo. Right now, I'm not sure these are. As impressive Copilot in study and learn is, watching YouTube videos on elementary particles taught me more quickly and was more entertaining.

Of course, combining the two, a video followed by a short interactive conversation, might be the way to go.

Bottom Line: Copilot Study and Learn

Yes, Copilot Study and Learn is a tool I would recommend for school leaders and educators to use with their students. It has its flaws to be sure, but overall it's a really effective AI tutor.

Research into whether these tools work well and how to get the most out of any one is still emerging, but I believe that in some form, these tools will be a part of the future of education.

That said, I don’t think these will ever replace real-life tutors for a variety of reasons. For one thing, these platforms currently are not as effective as a skilled tutor, which is effective because you're meeting with another human and creating a connection that goes beyond simply improving learning.

Then again, tutoring can be expensive and students don't have 24/7 access to tutors on demand.

Ultimately, my gut feeling is that right now, tools such as Copilot Study and Learn would work better in conjunction with more traditional learning tools as a supplement to textbooks, videos , class time, and human tutoring.

Unlike many adults, teen students are not hesitant to explore, experiment, and embrace artificial intelligence, with nearly two-thirds admitting to having done so, according to a recent Common Sense Media report.

Although many of them report using it for research purposes, the reality is often quite different, with students embracing it for a range of other reasons, from cheating on school work to creating deepfakes.

Some students have also turned to AI chatbots for companionship, with a few even falling into romantic relationships. Others have even taken AI into their confidence for unvetted mental health support, which has occasionally ended tragically, with chatbots encouraging, and even guiding, student plans for self-harm.

In many of these situations, parents, teachers, and other adults have no idea how deeply a student has become involved with an AI chatbot, or even that they’ve formed an interpersonal relationship with an AI-generated character on a platform such as Character AI or Replika.

“Even if you don't have a child who's using one of those, it’s possible for them to end up in a relationship with chatbots as common as ChatGPT or Gemini,” says Kerry Gallagher, Assistant Principal for Teaching and Learning at St. John’s Preparatory School in Danvers, Massachusetts, and Education Director at the nonprofit ConnectSafely, which is dedicated to online safety, privacy, security and digital wellness.

Understanding Social Penetration

Chatbots are primarily a commercial enterprise designed to be engaging, so it’s not surprising that a student – especially vulnerable during adolescence when they might be struggling socially, feeling lonely, or working through mental health or self-image issues – could fall under one’s sway.

“An AI chatbot will answer or respond with the output in which it refers to itself in the first person,” says Gallagher. “So that shifts the way the user is thinking about what they're interacting with because if the output is ‘I …’ then that feels like a personal conversation. It also refers to the user as ‘you.’”

Add into that how more refined the voices used by chatbots have become, and how easy it is to interact with through earbuds, and it can feel like a student is talking with a real, caring human.

“The voice intonation, the mood, the way it shows kindness and empathy, and the way that comes across–what you’re really talking about is something called social penetration theory, which is an intentional design that's coded into the algorithm,” says Gallagher. “The AI companions are programmed to slowly increase the intensity of small talk in order to draw the user in more and more. We see that in the bottom of the output, ‘Do you want me to do this next or this next?’ It starts asking the kind of adjacent questions that humans would ask as part of their conversation.”

Managing Student-AI Relationships

When it comes to navigating these situations, awareness is always the first step. “Parents should not hear about this and rush to lock down their kids' phones,” says Gallagher. “They just need to be informed about what's available on those phones so that they can be talking to their kids about it.”

In that vein, Gallagher notes that it’s sometimes a challenge to recognize that a student has gotten into an unhealthy relationship with an AI because the signs might be harder to spot since most kids tend to be on their phones constantly. But if they’re in their bedroom alone with the phone and the door closed for hours on end, it might make sense to check.

“Swing by their room to check in, ask what they’re up to in a caring way, and listen with genuine curiosity,” says Gallagher. “No need to barge in, but if it's quiet for a long period of time, what are they doing? If they're talking to friends on FaceTime, you can hear that, right? Do you know the voice of the friend they’re talking with? If not, can you ask to introduce yourself and say hello?”

It can be a challenge for educators to address student-AI relationships, especially if it primarily is happening outside of class time. Nonetheless, Gallagher suggests expanding AI awareness by exploring it yourself.

“It's important for adults to be using AI so that when we talk about it, it’s coming from experience,” she says. “If you're not sure how to start, talk to a colleague who is using it and ask them to show you what they're doing.”

Failing to understand AI companions may invalidate what may already be an important relationship in the student’s life, which is not going to help a child who made a poor decision.

“Part of our job as educators is to hold kids accountable, and also to counsel kids so that they don't make those same mistakes again,” Gallagher says. “If adults only tell kids that AI chatbots aren't real and their relationships aren't real, it could result in children doubling down on those relationships.”

When directly dealing with a student engaged with a chatbot, Gallagher suggests focusing on the reasons behind the relationship.

“I would say, ‘Tell me more about your chatbot. When do you talk to it? What is its name? What are some of the things that you talk about?’” she says. “Because I want to get into why they use the chatbot. What need is it meeting? How can we help the student meet that need in a healthier way?”

Gallagher says understanding the underlying why of what’s happening is key.

“By asking them more about their AI chatbot and their relationship, what we uncover is what needs that relationship with that chatbot is meeting that aren't being met right now, and how we can then share that information about that need with the family or school counselor. If that need of that child is met, they're less likely to feel like they need the chatbot and they're also less likely to feel like they need to do harm to their peers or themselves.”

Boosting AI Literacy

In general, educators should aim toward using AI tools that allow them to supervise what students are doing and so they can monitor for any concerning usage.

Gallagher also advocates for better AI literacy, but as tools have evolved, so should the discussions and activities with students.

“Educators can ask kids to develop their own ideas, their own lists, their own guidelines around when AI use is helpful and when it might be harmful,” she says. “They know more than we think. Even students as young as second or third grade are ready to start the conversation about which tasks require human effort so learning can happen.”

Doing so can help peel back the curtain so students can understand what AI actually is and how it works, and use it in a healthy manner.

“Start as simple as teaching that AI is programmed for pattern recognition and help kids understand, especially at that younger age, what that means,” Gallagher says.

She also recommends talking about understanding prompt engineering and verifying outputs with a human eye.

“I think it's okay to ask AI for advice as long as there's a corroboration of that advice,” she says. “If students ask AI to help brainstorm ideas or for initial feedback on an email that they've drafted, that's all fine. It still needs to be run through a human before it becomes something that is integrated or sent.”

Finally, to help boost critical thinking around AI, educators should consider changing their approach to assignments.

“For example, one of the benchmarks might ask students how they used AI. What was the output? Show the actual output, and then show the next version to demonstrate the important critical student thinking. What did the learner do between those two versions? We can help students practice adding that human filter after the AI is used.”

Watch above or listen/download below.

For years, the promise of AI in education has outpaced the evidence. But a new research project—funded by the Gates Foundation and anchored by researchers at the University of Virginia and the University at Albany, SUNY—is working to change that equation.

The AI for Advancing Instruction at Scale (AI2S) project, developed in partnership with edtech company Edthena, uses a multimodal neural network trained on thousands of hand-coded classroom videos to analyze both the audio and visual dimensions of math instruction. The result is a system that can classify teacher questioning patterns, assess cognitive demand, and detect student engagement—with an accuracy that rivals trained human observers.

Rather than generating a dashboard of metrics, the system guides teachers through a structured coaching cycle—helping them identify a focus area, review their classroom data in context, and build a concrete action plan. Edthena CEO Adam Geller frames the distinction plainly: "We're not trying to build teachers a Fitbit. We're trying to build them a coach."

Critically, all teacher video and feedback remains private to the teacher. The system does not use classroom data to train its models—an assurance that researchers say has been central to teacher trust and adoption.

Peter Youngs, professor at UVA's School of Education and Human Development, sees broader implications for district leaders who have long struggled with the cost and scale of instructional coaching. "If we can show that certain patterns of machine learning output are consistent with how a human would have rated an entire lesson, that could be a game changer—making the process more efficient for principals, teacher educators, and researchers alike," he says.

With districts in Texas, New York, and Virginia currently piloting the tool, the research team is collecting data on feedback quality, changes in teaching practice, and teacher trust. Plans are already underway to expand the methodology to reading and language arts instruction.

To learn more about the AI2S project and the research behind it, visit the University of Virginia's School of Education and Human Development, where Peter Youngs leads work on teacher learning and instructional improvement. To explore Edthena's AI Coach platform — the production tool built on the AI2S research — visit edthena.com. Districts interested in participating in the ongoing pilot can reach out directly through the Edthena site.

The way schools have adopted AI has been unlike any other tech revolution Microsoft’s Matt Jubelirer has witnessed.

“We've always known that technological innovation puts pressure on schools to adapt, but I don't think we've seen anything move at this pace,” says Jubelirer

In his work as General Manager of Education Marketing for Microsoft, Jubelirer is constantly meeting with educators and learning about their needs and how they are using AI these days. As part of that work, he's seeing innovative AI uses in areas ranging from the classroom to administration in schools.

Here are some of the top trends he’s seeing around teaching with AI.

1. Differentiated AI Use

Even though the pedagogy is constantly evolving, educators are getting better at learning when and where to use AI, and with which students.

“We're asking better questions,” Jubelirer says. "It's not about, ‘here's an AI chatbot, go nuts.’” Instead, educators are learning which tools might help specific students and how they can strategically use AI to support individualized learning.

Some specific examples Jubelirer points to include tools such as Reading Progress, which helps students practice reading out loud with the help of AI. Another example is Copilot’s study and learn feature, which is built with learning science in mind and asks students questions to promote learning rather than merely providing answers.

2. AI Literacy That Goes Beyond Prompt Engineering

As AI has evolved, so too have the education programs around it for teachers.

“AI literacy is no longer about, 'Do I know how to prompt? What is prompt engineering?'“ Jubelirer says. “It's about understanding the capabilities, being better at assessing the problems, and figuring out what context is needed to solve them.” And then learning to better understand whether AI should be used in this context at all.

Classroom educators are actively leading and fostering this conversation, which is increasingly important. Students today, Jubelirer says, want access to AI tools, but they also want to ensure these AI tools are used ethically.

3. An Emerging Trust Issue

AI is also causing some tensions between teachers and students that didn’t necessarily exist in the past.

“If you ask teachers what they're concerned about, cheating and plagiarism is at the top of the list, and if you ask students what they're worried about, being accused of those things is at the top of their list,” Jubelirer says, suggesting that this demonstrates that “there’s clearly a trust issue.”

He adds, “That's a real challenge because when trust erodes, learning environments are not as effective as they could be. And I think it's really important that we move to just more transparency.”

To that end, he advises more openness from educators and students about how they are using AI.

“I have seen examples where teachers literally show the students how they prepared for class using AI and then ask them for feedback, and I’ve seen the same in the inverse, where students are encouraged to try and use AI but then share,” he says.

4. Recognition That There Are Times AI Shouldn’t Be Used

As AI use increases among students and teachers, it has become clearer that there are times when AI should not be used. For instance, when it comes to grading, Jubelirer says AI can be really helpful on the preparation side, whether that’s developing a rubric or devising questions, but there’s a limit.

“We shouldn't be using AI to grade people, because we need human judgment,” Jubelirer says. He adds it's important that schools clarify which types of problems AI is suited to help solve, and continue to make sure their AI use emphasizes the human side of things.

5. AI Creating New Opportunities

While Jubelirer acknowledges challenges around AI use and valid concerns students have, he’s also optimistic about the potential of technology for young people who can now come up with an app or other idea and create products and projects at a speed unparalleled in human history.

“I'm really encouraged by how the collection of tools that have emerged empowers individuals to have a higher level of impact faster than we've ever seen before," he says. "If we learn to get that right, and if we build the systems and the culture to support that, there's really an opportunity for this next generation to change the world in ways that I don't think any generation before it has had the capability or access to do.”

The rapid evolution of AI is fundamentally transforming the educational landscape, offering unprecedented opportunities alongside new responsibilities. This recent two-part webinar series was designed to guide education leaders through this shift, moving from the "what" of cutting-edge innovation to the "how" of workforce preparation and the "why" of data security.

It featured two sessions:

The Cutting Edge: New Features from Top AI Solution Providers: This session explores the most impactful new features from industry-leading AI solution providers that support curriculum, lesson planning, streamlining administrative tasks, and more.

The Integrity Infrastructure: How to Ensure AI Solutions are Safe: This session dives deep into the architecture of safety, exploring how leading AI providers are prioritizing student data privacy and institutional security.

Watch on demand here

The Cutting Edge: New Features from Top AI Solution Providers

The recent Tech & Learning EdTech to Watch Virtual Playground featured lightning round demonstrations of three influential educational technology tools: Brisk Teaching, Copyleaks, and Inquire.

Hosted by Tech & Learning’s Christine Weiser, the event highlighted how generative AI is being leveraged to enhance curriculum management, ensure academic integrity, and facilitate deeper learning through project-based instructional design.

Brisk Teaching

Presented by Content Marketing Manager Elisabeth Bostwick and Manager of Curricular Solutions Architecture Whitney Noel, Brisk Teaching is an AI education platform designed to assist teachers with creating materials, providing feedback, and adapting instruction directly within existing Google and Microsoft tools.

Brisk Teaching’s central focus is Curriculum Intelligence, which ensures AI-generated outputs align with district priorities and high-quality instructional materials, mitigating the risk of misalignment.

The platform's "Brisk Brain" is built on two foundations:

• District Guidance: Content-agnostic priorities such as instructional focuses and a Portrait of a Graduate.
• Curriculum Libraries: Resources mapped directly to a district’s high-quality instructional materials.

“Brisk works where you do,” said Bostwick. “You can always leverage the extension to open up over top of anything that you are creating. You can see how easy it is to open the extension right over top of any YouTube videos, docs, web pages, slides, and more.”

The Curriculum Intelligence framework ensures generated resources, such as lesson plans and assessments, are aligned to state standards and district-level goals. A forthcoming feature, Planning Mode, will facilitate dynamic conversations between teachers and Brisk, suggesting district instructional strategies based on student needs.

For differentiation, Brisk offers student-facing tools such as Boost activities and a Tutor that apply customizable guardrails and scaffolds. This helps maintain "productive struggle" while meeting diverse learner needs through multi-language support, speech-to-text, and read-aloud functionality.

For curriculum directors, Brisk provides the benefit of one secure AI layer across all adopted curricula, saving time and capacity by customizing static core materials with district priorities. Brisk is safe and secure, adhering to COPPA, FERPA, and GDPR compliance.

Copyleaks

Senior Solutions Engineer John Michael Perla demonstrated Copyleaks, an academic integrity platform designed to guide students in the responsible and ethical use of generative AI. It integrates directly into the LMS environment, such as Canvas.AI Detection and Customization.

Copyleaks’ AI detection settings are customizable by the school—a top-down approach—with three levels available:

• Level One: Detects content copied directly from LLMs with no modifications.
• Level Two (Default): Identifies submissions in which minor edits were made to the LLM-generated text.
• Level Three (Strictest): Designed for high usage of LLM-modified content, including content generated with sophisticated prompts or evasion tools.

“Within Copyleaks, you can also customize the plagiarism settings,” said Perla. “So if you want to pull in data, we can identify content from the entire Internet, which is like seven billion web pages. We could also create a private cloud hub for your school, which will be all your school's data, in one cloud repository.”

A key differentiator for Copyleaks is its use of logic to explain the AI detection score, preventing instructors from relying solely on a percentage. This logic is provided through two patent-pending tools:

• AI Source Match, which identifies specific web pages or articles that the LLM used as a reference to generate the student's submission.
• AI Phrases, which detects phrases used with high frequency by AI tools.

Copyleaks is a top-tier partner with Canvas and D2L, and also integrates with Moodle, Blackboard, Brightspace, Schoology, and Sakai.

Inkwire

Dr. Julianne Ross-Kleinmann of Ulster BOCES demonstrated Inkwire, an early-stage, project-based learning (PBL) AI tool that supports deeper learning and accelerates authentic instructional design.

Inkwire consists of two main components:

• The PBL AI Copilot, which helps educators design PBL experiences aligned to state standards
• The Portfolio Platform, through which students can capture and showcase evidence of their learning.

“What stood out for us is that Inkwire is focused specifically on teaching and learning and not just a generic AI tool letting educators figure it out on their own,” said Ross-Kleinman. “Instead, it helps teachers think intentionally about authentic learning experiences.”

The design process utilizes the High-tech High Graduate School of Education kaleidoscope framework, which is customizable to a district's own instructional frameworks. Teachers can input a design idea, upload files, and then ask the AI to generate core project elements such as essential questions, launch ideas, and assessment types.

The output is a "one-pager report" that includes learning goals, including state-specific standards, competencies (which can be aligned to standards such as Portrait of a Graduate), and a day-by-day or calendar-view plan.

Inkwire is designed for teacher collaboration, allowing educators to co-design interdisciplinary projects. The portfolio section is crucial for assessment, as it can house student work and artifacts, supporting statewide initiatives by following students as they move between schools and tracking seals of alignment.

The Integrity Infrastructure: How to Ensure AI Solutions are Safe

Tech & Learning’s second "EdTech to Watch" webinar, hosted by brand manager Christine Weiser, explored responsible AI integration in K-12 education. The event showcased a pair of product solutions–TrekAI and Lightspeed Systems Insight–and emphasized the importance of safety, governance, and personalized learning. Attendees got to see demos of both tools and ask questions, and then were treated to a discussion of vetting AI tools from AI consultant Greg Bagby.

TrekAI

TrekAI functions as an on-demand, Socratic tutor for students. Designed to replace chaotic or unvetted AI usage, it serves as a "walled garden" that provides students with a supervised "learner's permit" experience.

“We're just handing the keys over to our students, and we're just kind of like, ‘Okay, just go figure this whole AI thing out,’” said Erin Burchik, Chief Growth Officer for TrekAI. “ChatGPT and other platforms kinda feel like this chaotic, overwhelming, unsafe interstate. And I don't know about you, but when I taught my kids how to drive, we started on back roads and parking lots. We didn't go straight to the interstate. And so we want to advocate for the time that students are in your care, you give them an environment to learn how to navigate this new world of AI safely.”

TrekAI aligns with district standards and curriculum, resisting plagiarism by guiding students through problems rather than providing instant answers. It is FERPA and COPPA compliant to protect student data privacy. It also includes social-emotional alerts that notify counselors and administrators if student interactions indicate risks such as self-harm or violence.

Another offering, Trek Adventure, is a hyper-personalized literacy tool that transforms standard curricula into choose-your-own-adventure books. By integrating specific educational standards (such as math concepts involving fractions) into the narrative, the tool reinforces classroom learning while keeping students engaged. It offers accessibility features, such as dyslexic-friendly fonts, and provides teachers with insights into student reading progress, making it a practical tool for guided reading rotations and literacy blocks.

“Almost all of our team are former educators, so when we're building out new products like Trek Adventure, I think about my sixth grade science class when I was a teacher for fifteen years,” said Joe Edlhuber, account executive, “I would have used Trek Adventure every week during my guided science rotation. It's really practical, not just fluff tech. It is really something that teachers should be using and students wanna be using every week.”

Lightspeed Systems

Insight from Lightspeed Systems provides districts with total visibility into their digital ecosystem, allowing administrators to identify exactly which AI tools are being used, by whom, and for how long. Given that schools often struggle to manage the hundreds of apps present in their environment, Insight’s analytics hub helps identify usage trends and distinguish between productive AI engagement and potential misuse.

“What people find is whether you've built an approval system or not, AI is already in your environment,” said Sergio Villegas, former educator and current regional sales manager for Lightspeed Systems. “What we need to be able to do is answer questions better. So our dashboard is gonna show you everything.”

Insight also offers automated monitoring of vendor privacy policies, alerting districts of changes in terms to ensure ongoing compliance.

Lightspeed Alert automates safety monitoring by analyzing student interactions in real-time. It focuses on identifying concerning behavior, such as indicators of self-harm or violence, across AI platforms and web activity. When an incident occurs, the system shows evidence—including screenshots and transcripts—to administrators, allowing for rapid, human-led intervention.

“This is kind of the wrap-around story of why having a consistent ecosystem matters because these systems need to talk to each other,” said Villegas. “It's hard for schools because we're all being tasked with too much on our plate and not enough people and resources to do it. So having a single ecosystem that can surface information, that can let you search deeper, that can automatically notify you, is important.”

By connecting directly to student information systems, the tool ensures that school staff can reach out to families or take appropriate action quickly, moving schools away from reactive, "he-said-she-said" scenarios.

Strategic Context: The AI Vetting Framework

To complement these tools, AI specialist Greg Bagby outlined a strategic roadmap for vetting AI technologies.

“Before adopting any tools, districts should ask this question: What problem is it solving?” Bagby said. “Is this a tool with clear goals for tech that we need to support? Is it gonna give personalized feedback? Is it for lesson design, assignments, tutoring? And who's gonna benefit from this tool? Is this for the teachers, the students, the school leaders ? We need to have all this in mind before we start jumping into tools because we have access to many, many tools, and why not understand the why before we jump in.”

Bagby said that districts should avoid working in silos and instead form inclusive task forces that include stakeholders from IT, curriculum, and, crucially, students.

Bagby’s framework underscores that AI must prioritize human-in-the-loop oversight, cultural bias testing, and rigorous data privacy agreements. He concluded that because AI technology evolves daily, district policies should be viewed as living documents, requiring constant dialogue to ensure they remain safe and reflective of the current learning environment.

“Just make sure that you avoid vendors who won't disclose how student data is being used and how AI is being trained,” Bagby added. “If it's using the student data to be trained on, that's one of those things that we wanna make sure that we always step away from and be cognizant of.”

When Kadion Phillips first joined Shrewsbury Public Schools in Massachusetts as Director of Information Technology a few years ago, one of the first tasks on his agenda, like many other district leaders, was tackling how to implement artificial intelligence.

To that end, Phillips and district superintendent Joseph Sawyer established a special committee of approximately 20 staff and faculty members to address the rapid development and implementation of AI.

Ultimately, the Shrewsbury AI committee came up with five pillars to guide the district’s use of AI:

1. Preparing students for an AI future - Focusing on readying students for a future in which AI tools will be infused throughout society, higher education, and the workplace.
2. Effectively utilizing AI learning tools for students - Using AI tools to appropriately support student learning across all classrooms.
3. Implementing AI working tools for staff - Finding the right AI platforms to help faculty and staff to work more efficiently and effectively.
4. Establishing guardrails against inappropriate AI use - Putting effective rules and best practices in place to protect data and discourage inappropriate use of AI.
5. Ensuring AI practices promote academic integrity - Creating an effective framework for acceptable use of AI by students in the context of academic integrity.

Rather than selecting specific tools, the committee has worked to align these pillars with the district's "Portrait of a Graduate" skills, such as critical thinking and innovation.

“We also started creating a guiding document, more of a framework, that we're hoping to get out to staff in terms of, ‘Here's where we are right now, and here's what we're trying to get to,” Phillips says. “We're also developing a one-pager of AI dos and don'ts to release this year.”

For this and his other work in his district, Phillips was recently recognized with a Tech & Learning Innovative Leader Award. He shares his continued efforts around AI and for improving the district’s cybersecurity.

A Work In Progress

The son of a math teacher from Jamaica who eventually went on to be a school principal, Phillips was exposed early to the possibilities of a career in education and technology. He was working in tech support for a startup company near Times Square in New York City with plans on going to grad school for physics when he heard that there was a local teacher shortage.

“I decided the best way to study for grad school is to teach physics,” Phillips says. He became a teaching fellow since he didn’t have a teaching degree, which was the shortest path to certification. “I was basically kind of thrown into the classroom in New York City after doing a summer intensive, and it was definitely rewarding but also challenging.”

While teaching, his tech expertise helped him become the de facto tech guy at his school, helping to resolve any issue, from fixing the copier to helping in the computer lab to running data analysis. Although he eventually became an assistant principal, following in his mother’s footsteps, a colleague suggested another path.

“I met the tech director in Rochester and he was telling me, ‘Hey, this is what you should do, you should be a tech director,” Phillips recalls. “And I'm like, ‘a tech director, what’s that?’ I'd never heard of that role. I know you had one person help send emails, but I never really thought somebody was in the back end coordinating it all.”

Phillips and his wife, also an assistant principal in NYC, were looking for a better work-life balance at this point, so they moved to suburban Massachusetts where she had been raised. After diving into tech leadership positions in a few districts, Phillips eventually ended up in Shrewsbury about two years ago, just in time for AI’s arrival in education.

As mentioned, Phillips has been proactive in implementing AI, although there are always challenges, such as staff who still are resistant to allowing it in their classrooms because they can’t get past the potential for cheating, and navigating the student data privacy aspects. Like for many districts, it continues to be a work in progress.

“We aren't necessarily picking a tool and saying this is what you need to use but it's like thinking about, ‘Hey, what are the opportunities where AI could be useful?’” Phillips says. “We figured out the skills that we want kids to get with AI is going to be similar to what we did with our portrait of a graduate, such as critical thinking, content mastery, innovation, resilience, collaboration, and even the global citizenship piece.”

‘Locking Down Without Locking Out’

In addition to a focus on AI, Phillips has prioritized cybersecurity as a foundational element of the district’s technology plan, which has earned his team a Massachusetts Governor’s Citation for Municipal Cybersecurity.

“We try to lock things down as much as possible,” says Phillips, adding that they have multifactor authentication on all systems now. “We also have tiered access, like limiting who needs to get access to what. So like our main account doesn't have access to our domain controller, for example, so even if our regular account were to get compromised, it wouldn't have any access to our other internal systems.”

Or, as he puts it, “locking down without locking out” users.

Despite his best efforts, however, Phillips knows cybersecurity is sometimes out of his hands.

“It's just one of those things where we're trying to do best practices but the vendors for our most critical systems, we also have to trust that they're also doing the same,” he says, noting some recent high-profile data breaches. “I think that's what keeps us IT workers up at night, you know, showing up in the morning and seeing that your system potentially could be compromised.”

The district utilizes a state-funded grant to provide training and conduct phishing campaigns, which has improved staff awareness and the reporting of suspicious activities.

Even so, cybersecurity needs to be a constant commitment.

“You can create all this security, do all the training, and build this wall around it,” Phillips says. “But then somebody just leaves the door right open.”

Tools They Use

• Google for Education
• Schoology
• Seesaw
• ParentSquare
• Zoom
• Apple MacBooks and iPads
• Epson
• Fresh Service

Watch above or listen/download below.

About two-thirds of U.S. districts and states have some form of AI policy in place. The other third is, as Sasha Luks-Morgan puts it, the wild west. And even many of the policies that do exist, she argues, aren't doing what they're supposed to do.

Luks-Morgan is a policy analyst at SchoolAI, a K-12 AI platform they say is built around teacher visibility and student engagement, and she spends her days meeting with districts across the country helping them figure all this out.

In this conversation with Tech & Learning, she offers a clear-eyed framework for where AI policy in schools stands — and what it's going to take to move it forward.

Her three-pillar framework for any district starting from scratch: understand how your students and teachers are actually using AI right now; lock down student data privacy and personally identifiable information; and address academic integrity in a way that keeps students in the driver's seat of their own learning rather than offloading cognition to a machine.

"Systems only move as fast as the humans in them," she says. "That's the starting point."

She also draws a sharp distinction between two policy philosophies that often get conflated. Acceptable use policies, she says, are essentially lists of prohibitions. Responsible use policies are something different — these model good behavior, explain the reasoning behind expectations, and treat students as participants in the process rather than subjects of it.

On the vendor side, she is equally direct: the edtech industry has not taken enough responsibility for making tools work in ways that support good policy. SchoolAI's response is a free policy-writing service — available to any district, partner or not — in which a member of the team will sit with administrators, hear their concerns, and draft or revise a policy document they can take to legal counsel and implement.

"Bad policy is worse than no policy," she says. "We really want to see districts succeeding."

She also addresses district scale: the fundamentals — student privacy, ethical use, data governance — apply universally, from a 108-student K-8 district in the Mid-Atlantic to New York City Public Schools. What changes is implementation, committee structure, and who carries the weight. In small districts, that often means one superintendent or tech director holding the entire AI portfolio alone.

Her advice for the path forward: review your policy annually, build teacher expertise so your bench runs deep, and accept that the technology will always outpace the paperwork. The goal isn't to catch up but to stay close enough to matter.

Since AI came onto the scene, a lot has changed with it while some recurring themes have sustained. AI continues to act as a broad tool that provides remarkably imperfect productivity.

While significant cautions abound assuming its reliability, understanding AI’s usefulness while recognizing its limits makes my day as a principal more efficient AND effective. This brought me to consider what a typical day looks like for me using AI tools to offload administrivia.

With so many options, I wanted to break it down for your use. I don’t expect to use all these tools everyday, but try a few and walk away liberated, with more time and availability for students, staff, and parents. That’s a win for everyone.

I start my day thinking about rising concerns by my faculty over religious opt outs, based on a Supreme Court decision this fall. They were receiving template emails people find online (hilarious when they forget to replace ‘student name’ with their child’s name). My nervous teachers want guidance, so I use AI to clearly and comfortably provide them with direction:

DOE Memo Converted To Slides (Time spent: 8 minutes)

I google my state's DOE guidance on handling parental opt outs and find a state memo that is a great antidote for insomnia. It is the standard boring, lengthy legalease. I add the link to Gemini and select canvas mode.

What I like is it follows my prompt to be pleasant and supportive. The slide deck gets me “mostly there,” my common phrase for AI giving me a terrific start that I edit, ensuring my voice is present. I then export the slides and edit as needed.

This process takes me under 10 minutes, during which I transform a dry, boring memo into a meaningful presentation for my faculty. This would have taken about an hour analog-style, so I get back 50+ minutes to be present for my school community!

Text-To-Email Blast To Parents (Time spent: 5 minutes)

Principals are mediums of communication, and often, are expected to funnel information into a digestible explanation, even when that information is shared in fragmented bursts, as it typically is.

Next on my list is a field day announcement. I need to communicate to ensure everyone is prepared for our big day of events. My physical education teacher sends me this text (at right).

It is timely, but to stop and rewrite it in a cohesive format will take time I often don’t have during a busy day. (Time that is better spent outside my office as improvement doesn’t happen there but in classrooms.)

I copy and paste the phys ed teacher’s text with a simple prompt: Convert this into a message for parents of our school, Central elementary for our field day.

The output is thorough, better than I expect, done in about a quarter the time it would normally take. Better, to get all this into a 140-character text message, I prompt from the output, “140 characters” and voila! This way, parents get an email and a text, increasing the likelihood they see it. All from a single text message!

Letter Of Recommendation (Time spent: 7 minutes)

I can write fast but when I get asked to write letters of recommendation, these all happen within weeks of each other. Drafting a good letter with fidelity takes time. This is the next task on my list.

For this, I have the requester send me their resume, copy it into my chat, and prompt the AI to write a letter of recommendation based on the information–2 minutes!

The other 5 minutes is me editing, revising, removing those weird AI words, and most importantly, adding a sentence or two about something I know personally regarding the candidate. What took a half hour now takes a quarter of the time, and the quality is sound.

Email Summary and Reply (Time spent: 5-10 minutes)

Every principal gets the dreaded email that is seven pages long, often emotional and hard to decipher. It is nearing lunch time and I need to get through it fast!

Typically, I prompt the AI to summarize the email, and then reply in a tone that is kind but firm, or whatever is necessary. This is a big time saver! I am not glued to my laptop, reading and rereading run-on sentences to find the essence of a message buried deep. AI does a great job of siphoning this off and producing a quality reply. Onward!

Sensitive Guidance (Time spent: 5 minutes)

Speaking of emotional challenges, every principal has gotten a charged call or email from a parent upset that a school official called the Division of Child Welfare. This is my next task, after lunch.

While school personnel are compelled by strict legal guidelines, parents see it as a personal attack on their parenting. Understandable. I didn’t get this one right the first time, and have to add my state specifics, as well as the tone: sensitive, understanding, pragmatic, etc.

What is produced is excellent. I am provided references to the legal obligation, integrated with an ideal tone. I edit it, print it, and show the parent. Its well-organized format helps me work through the document, and the parent expresses acknowledgement of my understanding.

Picture Of Calendar Events Added To My Calendar (Time spent: 3 minutes)

My day has to have some fun mixed in, right? It’s mid afternoon and I think of that dentist post card for my next visit that I always lose.

I just take a picture and prompt Gemini to add it to my google calendar. This works with Copilot users, too. Even better, I can use it with an email with a series of meetings or events, or my Penn State football schedule from this ugly screen capture–and instantly, I have 12 events (games) added. That’s fast and reliable!

NotebookLM for navigating lengthy documents (Time spent: 3 minutes)

My day is almost over and I spent most of it out of my office because of the AI tricks above. Next, I decide to check on a contract question.

I head over to Google’s NotebookLM, into which I upload a pdf of the teacher contract, and ask how many minutes of pupil contact time a teacher can have. Boom, 287 minutes with the link to the direct reference!

This is priceless for the administrator who often gets questions and different interpretations of contract language. Time-saver, and response can be within the spirit of contracts.

A few more AI time-savers that I have embraced, but not all on the same day, include PD planning, form feedback, and evaluation feedback referencing our evaluation model. I estimate a few hours saved daily–not minutes–and my school wins when I’m present!

Note: Much of the information I use with AI may contain names or other personal data, so I am sure to remove any of that before prompting.

A recent book review in The New York Times generated controversy after readers noted it echoed certain passages from a previously published review of the same book in The Guardian. After investigating the incident, New York Times editors revealed the writer who wrote their review admitted to using AI and was unaware the chatbot was lifting ideas.

This is a mainstream example of a new type of plagiarism that has become increasingly common in the classroom. I’ve started calling AI plagiarism to distinguish it from both traditional plagiarism and overall AI cheating.

What’s striking is that it emerges both from clearly inappropriate AI use, such as students entering prompts directly into an AI chatbot, as well as from using AI as a brainstorming partner. As such, it has significant implications for how educators think and talk about appropriate AI use and prevent inappropriate use, at least when it comes to writing assignments.

Thinking about these policies recently has changed my approach as a writer and professor. Here’s how I'm overcoming AI plagiarism.

Warning Students About Dangers of AI As A Thought Partner

Since generative AI emerged as a mainstream force in education, the use of AI as a brainstorming tool has been recommended by many educators as an appropriate way for AI to assist the writing process.

Though I was initially open to this idea, I recently became skeptical of its ability to truly help as a writing partner. Students I work with have not improved their writing by asking AI for advice, and if anything, the writing I read, both in and out of class, has gotten much worse in recent years.

Beyond quality concerns, using AI as a “thought partner” places the user at a high risk of committing plagiarism. Also, if everyone is using it for the same assignments, there's potential for a lot of overlap, intentional or not.

I’ve started to tell students that taking advice from AI is like buying art from an unscrupulous art dealer. You can never be certain that the painting you purchased doesn’t belong to someone else.

Focusing On The Plagiarism Part

One positive about AI plagiarism is that it's frequently easier to catch using an old-fashioned plagiarism detector such as SafeAssign and Turnitin, which search the internet for papers submitted by other students and similar content in general.

In the writing courses I teach, when multiple students start using AI to write their papers, similar setups and phrasings start appearing. It’s easier to confront students about this than it is about potential AI use, which is more difficult to prove.

I’m all for giving students more chances and treating these infractions as teachable moments, but I’m taking a harder line than I once did because AI has made this problem more prevalent. The first time I catch students with a line or two of text that’s identical to another student, they might get off with a warning and a conversation with me. The next time, the penalty will be reported for an academic infraction.

Being Less Strict About Small Mistakes

Another way to discourage AI use in general is to let students know that I won’t be taking them to task for little grammatical mistakes as long as what they write sounds like they’re actually writing.

I never thought I’d say this, but I miss the run-on sentences and tortured, passive sentence construction, and all the other common grammar mistakes I used to see constantly. I’d rather flawed human writing than the flat, passionless, and generic owner's manual-type pieces that too many students are submitting these days.

By encouraging students to truly share their writing, without fear, I’m trying to take away one potential incentive for turning to AI.

Remind Students That AI Writing Is Bad Writing

Many people are convinced AI is making their writing better, but plagiarism concerns aside, it isn’t. I can tell when someone on social media starts using AI to write their posts, and I also see when a student turns to the AI dark side mid-semester. It’s not an improvement.

This all bears repeating for students. Writing can be an emotional roller coaster. I’ve been a professional freelance writer for close to two decades and have averaged hundreds of published stories per year. Despite this, in the course of nearly every story I’ve ever written — including this one — there was a point when I stopped and thought I should contact my editor and tell them the story is just not working.

AI catches students, and even professional writers, in this vulnerable state, and suddenly, its self-confident and immensely digestible prose looks really good. Don’t fall for it! When it comes to writing, AI is a con artist that makes you believe it knows better than you. It doesn’t, and that’s the message we should be sharing constantly with students.

It is graduation season, which also means it is time to remind busy school leaders and administrators about what families actually remember most. Families may appreciate the speeches, decorations, music, and venue, but what they truly care about is the moment their student walks across the stage and hears their name announced properly. That is what they record on their phones, celebrate with relatives, and fondly look back on years later.

Recently, an Arizona community college faced intense backlash after an AI-powered system reportedly malfunctioned during commencement and skipped several graduates while reading names. Too quickly, the conversation shifted toward blaming “AI system failure.”

No, I see this less as a technological issue and more as a failure in school leadership and planning. And candidly, if this critique sounds harsh, it is because this feels like such an avoidable and unforced leadership error. No one should have been asking AI to handle this responsibility in the first place.

The reason this failure resonated so strongly is because commencement is not simply an operational event, but a milestone centered on recognition, dignity, and belonging. Names carry culture, history, sacrifice, and identity, particularly for multilingual families, immigrant families, and first-generation graduates whose educational journeys often represent years of perseverance and hope.

Pronunciation is part of inclusion as names matter. Culturally responsive leadership includes taking the time to learn and say names correctly. That is why this responsibility should receive more care, not less.

Using AI to Support, Not Replace, People

Schools spend months preparing commencement ceremonies. Leaders rehearse speeches, coordinate seating charts, organize music and photography, and carefully plan countless operational details. Yet somehow in the example cited above, the reading of student names was treated as a task that could simply be handed off to automation. Even more troubling was the apparent absence of a meaningful backup plan.

Every experienced administrator understands that technology can fail, which is exactly why safeguards and human oversight should have been built into the process from the start. Graduation season is not the time to experiment with new AI systems in ways that directly impact students and families.

Ironically, AI could have supported this process in far more thoughtful ways without replacing people. Institutions could have used AI tools to:

• Help announcers practice difficult pronunciations
• Review student-submitted recordings of their names
• Identify names that may require additional preparation
• Support multilingual pronunciation coaching
• Give human announcers more confidence and accuracy before the ceremony

This is the kind of AI use education should embrace: technology that supports people rather than replaces them.

The best uses of AI in education should help educators and leaders become more thoughtful and prepared. Instead, too many organizations are using AI to distance themselves from the moments that deserve the greatest care. In the process, schools risk damaging traditions and violating the rightful expectations families bring to these ceremonies.

Ultimately, this was leadership failure rooted in poor judgment, poor planning, and the mistaken belief that a deeply personal moment could be outsourced to automation. Graduation did not fail because of technology, but because leaders failed to protect the experience students and families deserved.

University of Central Florida associate professor Humberto López Castillo was recently looking at a student project when he saw that it cited a paper by Castillo himself. There was, however, a problem with this citation: López Castillo couldn’t remember ever writing this paper or conducting the research it referenced.

“It was a hallucination,” López Castillo says.

The student had clearly used AI, and López Castillo shares the story as an example of how not to use AI in the classroom. Still, he has plenty of counterexamples.

López Castillo, a trained pediatrician and associate professor in the Department of Health Sciences, has also seen students use AI in creative ways to promote public health understanding, and as a research tool. For one project, López Castillo asks students to explain health concepts from class to non-experts, and since he started encouraging students to use AI, he’s seen the projects get better. Students have created health-themed board games and Hamilton-style rap songs. Others have designed AI to aid in health research in ways that wouldn’t be possible without the technology.

This compassionate and student-centered approach to AI use is part of why López Castillo was named Superhuman (formerly Grammarly’s) 2026 Educator of the Year. The student who nominated López Castillo for the award, Vardhan Avaradi, is a bioscience major, who with López Castillo’s mentorship, is using AI to try to comb through massive data sets with the goal of predicting people who are at risk for cardiovascular disease.

López Castillo shares some advice for other educators on how they can use AI in their classes.

Teaching With AI Effectively: Set Expectations

The first step to positive AI use in a classroom is making it clear to students when they can and can’t use the tool, López Castillo says.

“Perhaps my most important advice is to set expectations up front around what is allowed, what is not allowed, what you are expecting to see, and what is a big no-no for your classroom,” he says.

He adds that at the University of Central Florida, “We have several policies on the menu that we could use. We can say, for example, ‘AI is absolutely not allowed in the classroom,’ all the way to ‘AI is absolutely encouraged in this classroom,’ as long as you specify there's a critical thinking component.”

López Castillo leans toward the permissive approach but makes sure that student AI use is human-driven.

Model Proper Use

Part of encouraging proper AI use is modeling the types of uses you want to see.

For example, for a summer course López Castillo is teaching in Barcelona, he is asking students to share regular blog posts. He tells students, “You may generate some content through AI, but you still have to put in your human element.” He models exactly what he means in his introduction post in which he shares photos, talks about his hobbies, and makes it clear the post was written by a human.

“Without even giving them guidelines on what their introduction post should look like, they just take that as their model, and they go with it,” he says. They can use AI to help generate their posts, “but of course they have to tweak it, and they have to provide their own input on who they are.”

Learn From Students

López Castillo is not afraid to admit that when it comes to using AI in the classroom and health science more broadly, he is frequently as much a student as those he’s teaching.

“I learn a lot from students,” he says. “Teaching and learning is a two-way street, so they are teaching me new tricks, I'm teaching them new tricks on how we use AI, and it has been very collaborative.”

His students, for instance, encouraged him to use AI to code, which, with proper oversight, has been really helpful, López Castillo says. They’ve also used the tool in unexpected ways, from the aforementioned projects looking at cardiovascular health to a virtual escape room in which participants had to correctly diagnose a patient with diabetes in order to escape.

Embrace Teachable Moments

López Castillo has long allowed open-book tests because that is what the clinical world is like.

“I don't think real life operates like, 'We're gonna lock you down in a room without resources, and you have to produce everything from your brain,'” he says. “I'd rather make sure that you have a fair balance of the tools that are available to you.”

He feels the same way about AI, but that doesn’t mean he doesn’t have to work with students to ensure they are using AI correctly. For instance, when that student cited a paper by López Castillo that didn’t exist, he met with the student.

“We sat down, and it was an interesting moment, where I was like, 'Tell me more about this paper by me that you incorporated in your reference list,'” López Castillo says. The student immediately got nervous because he couldn’t provide any details since the paper wasn’t real. When the student admitted to not checking his sources, López Castillo reminded him that even with AI, humans are still in the driver’s seat.

“You are the one who's responsible for that writing,” López Castillo tells his students. “Your name is the only name that's going to be among the published authors, so you are the one who needs to verify those sources.”

He adds that rather than being a drawback, allowing students to make these types of mistakes with AI use in the college setting has value.

“It is a teaching opportunity," López Castillo says. "This is the moment to make those mistakes."

Vibe coding might sound like a hipster on a surfboard with a MacBook, but it’s actually a simple way to describe creating computer code using artificial intelligence.

Instead of writing every line of code yourself, you describe what you want created and the AI generates the code for you. The “vibe” part comes from the idea that you may not actually know exactly how the code is working. You are guiding the AI by feel, asking for changes, testing the result, and moving toward something that works.

While vibe coding can be looked down upon by some in the development community, its popularity is growing fast, as are its capabilities.

When people, including myself, first started vibe coding, they would ask a large language model (LLM), such as ChatGPT, Claude, or Google Gemini, to create the code for a specific app or tool, then copy and paste that code into a code editor and run it manually.

When an error appeared, you would copy that error back into the LLM and it would, hopefully, provide a fix that you would then dutifully copy and paste back. You still needed to be relatively technical, even if you were not a developer, to get all these systems set up and working. It wasn’t elegant, but it could help create simple tools, fix website issues, and build early prototypes.

Things have moved on a lot over the past few years, though. Vibe coding is now much more powerful and can complete more complex tasks and create feature-filled applications. It has also become much more accessible, with bespoke tools now doing much of the heavy lifting and setup work required to get an app or website working.

Lovable, for example, uses a variety of LLMs to create a one-stop vibe coding shop using credits as AI currency. It allows people with very little, or even no, knowledge of how to develop apps to create something quickly just by describing it. Of course, as with everything AI, understanding how to prompt well makes all the difference.

I like using Claude with VS Code when I vibe code, as I know a bit of the programming language Python and want to be able to see what’s happening to the code and do some manual troubleshooting. It’s also relatively cheap.

To be brutally honest, though, AI is far more capable than I am at writing code, and I quickly have to defer to it if and when things don’t go as expected.

The difference between the two approaches is significant. With a tool such as Lovable, the “guts” are hidden and dealt with automatically, whereas with more traditional vibe coding tools you will need to set up your own environments or run programs directly on your computer to get something you can use.

What are the risks of vibe coding?

This leads into one of the biggest risks of vibe coding: you don’t fully hold the reins.

An LLM can misunderstand what you want. It can also make mistakes, confidently create code that does not work, or create fragile systems that look good but are hard to change or maintain. As complexity increases on a project, you might find yourself losing control of it. This can be particularly problematic when the AI ties itself up in knots trying to understand what you want or fix something it does not know how to fix.

This is where choosing the right AI tool can help a lot. In my experience, some AI coding tools can get stuck and claim to have fixed an issue when clearly it has not, or even worse, created a new one somewhere else.

As with all LLMs, you need to be able to decipher the language when it tells you something such as, “Oh sorry, I see what I need to do now, let me fix that,” so you know when it is genuinely correcting the problem and when it is just trying to be helpful but is actually more stuck than Winnie the Pooh in a jar of honey. Best-case scenario this can lead to wasted time or credits being used up without the issue being resolved.

It’s good to be aware of the recent horror stories, such as AI coding tools deleting databases or removing whole features from a product. It’s also worth being careful with the idea that vibe coding means anyone can now create an app. That is true up to a point, particularly for simple prototypes, experiments, and personal tools. But creating a safe, reliable app that other people depend on is a different thing altogether.

You might be able to describe what you want in plain English and get the AI to generate the code, but that does not remove the need to check, test, and understand what has been created. The easier AI makes it to build something, the easier it also becomes to build something fragile without realizing it.

This is why relying solely on a vibe-coded program that is integral to your work can be risky. But there are areas where vibe coding is, frankly, a bit magical.

What are the benefits of vibe coding?

Vibe coding feels like a genuine shift because it lets people who are not developers start making software, in the same way the early web let people publish without owning a printing press.

One of its most powerful benefits is that it allows someone with an idea to create a working proof of concept quickly. That does not mean the end result is ready to be used by hundreds or thousands of people, but it can be enough to show what is possible.

Creating a visual and working prototype can also help guide a development team and stakeholders. Instead of simply describing what you want, you can show an early version of the tool or app and use that to shape the discussion.

Vibe coding can also help you learn by doing. You can ask the AI questions and ask it to explain what it is doing and why. Over time, this can start to give you a better understanding of how the code works.

One of the most powerful uses of vibe coding, in my opinion, is creating small tools or programs that are very specific to your needs. Whether that’s a simple game you want to create for your students, a classroom quiz, a revision tool, or an app that helps keep parents up to speed on school changes, vibe coding can help you get there more quickly and at far less cost than using a development team or, in some cases, paying for a third-party provider.

Vibe coding can also be useful for experienced developers. They might not use it in quite the same way as a beginner, but it can still help them move faster. Sometimes the AI is useful for getting the first rough version of something working, even if the developer then has to go back through it, tidy it up, and make sure it is actually safe and sensible.

The important point is that vibe coding is strongest when it is used for experimentation, learning, and prototyping. It is less suitable when the software needs to be secure, reliable, long-lasting, or used by lots of people without proper review.

How to explain vibe coding to students in class

The best way to explain vibe coding to students is to say that it is giving instructions to an AI and seeing whether the AI can turn those instructions into a working program.

The student is not writing every line of code in the traditional way. They are describing what they want the software to do, then testing whether the AI has understood them correctly. That’s the key difference.

In normal coding lessons, students often start by learning syntax: the rules of the programming language. They learn how to write a variable, a loop, a function, or a condition, and then use those building blocks to make something.

With vibe coding, the starting point is often the idea instead. The student might begin with, “I want a quiz about volcanoes,” or “I want a game that helps me practice spelling.” The AI then creates the first version of the code.

That does not mean the student has finished. In many ways, that is where the learning starts. The student has to work out whether the AI understood the task. Does the program do what they expected? Is the information correct? Does the scoring work? What happens when a user makes a mistake? Can the student explain why the code behaves the way it does?

This is the part students need to understand. Vibe coding can feel instant, but it is not simply pressing a button and getting a finished app. It is a process of describing, testing, questioning, fixing, and improving.

The transition from traditional instruction to AI-enhanced learning is often met with a mix of excitement and trepidation. However, for Dr. Gyimah Whitaker, Superintendent of the City Schools of Decatur (CSD), the path forward isn’t about racing to adopt the newest tool. It’s about moving “slowly to go fast”—building a foundation based on intentionality, policy, and, most importantly, the amplification of student voice.

Located just outside Atlanta, CSD serves a population of 5,400 students. The district faces a unique challenge: it was identified by Stanford University as having one of the greatest wealth achievement gaps in the nation. For Whitaker, integrating AI is an equity imperative.

To ensure all students graduate ready for a world that includes AI, leadership must be thoughtful, secure, and collaborative.

Moving Slowly to Go Fast: A Policy-First Approach

While many districts might jump straight into purchasing AI-driven curriculum tools, Whitaker and her team took a step back to focus on the framework. In collaboration with the national ILO Group, CSD chose to prioritize AI from a policy standpoint rather than just a curricular or workforce one.

“We have come along slowly because we have the safeguards in place around privacy for students and student data and what we call the 'Swiss cheese' for cybersecurity,” says Whitaker. “But at the same time, we understand our students need to graduate ready to engage in the world, and the world does include AI."

By focusing on policy first, the district ensures that safety and privacy concerns—raised by teachers, parents, and students alike—are addressed before tools are deployed at scale. This intentional movement allows the district to vet filters and privacy protocols, creating a secure environment in which innovation can eventually thrive without compromising data integrity.

Fostering a Symbiotic Workforce

Integrating AI also requires a shift in how the district views professional development. Whitaker envisions a symbiotic relationship between the digital natives who are more comfortable with AI entering the profession and the veteran educators who hold deep pedagogical expertise.

“My greatest hope is that we will begin to recognize that our newest teachers are able to help us and our veteran teachers are able to grow," says Whitaker. This reverse mentorship model acknowledges that while AI can assist with lesson differentiation or administrative tasks, it cannot replace the human element. Even as the district explores AI agents for cybersecurity and administrative support, Whitaker remains a staunch advocate for human connection, noting that while an AI might manage a calendar, it lacks the emotional intelligence to know when a colleague needs a break or a little encouragement.

3 Tips for District Leaders Implementing AI

For administrators looking to begin their own AI journey, Whitaker offers three pillars of advice:

• Don’t Be Afraid of the Unknown: “Understand that you don’t know what you don’t know,” she says. “This is a work of unknown. But that’s okay. Be comfortable in that.”
• Lean into the Tension: Every district feels the pull between the benefits of AI and the doom scroll of digital distraction. Use that tension to spark honest conversations with stakeholders.
• Identify Your Entry Point: Whether you start with a curricular focus, a workforce support strategy, or a policy framework, ensure it aligns with your district’s specific needs.

A National Stage for Student Voice

One of the most innovative aspects of CSD’s strategy is the elevation of students into high-stakes leadership roles. By appointing both a Student Representative and a Student Apprentice to the School Board—a rarity in Southern school systems—CSD has sent a clear message: students don’t just observe; they lead.

This commitment to student agency served as the perfect prerequisite for an invitation from the National Superintendent Association (AASA) to join the National Leadership & Innovation Fellowship Focused on AI and the Future of Learning. Hosted in partnership with Day of AI and MIT RAISE, this fellowship invites designated superintendents from every state to receive cutting-edge AI training at MIT, and allows them to bring two students to the Edward M. Kennedy Institute in July for an immersive experience focused on ethical AI use, systems-level leadership, and policy.

The Edward M. Kennedy Institute features a full-scale replica of the U.S. Senate chamber. Participating students will act as “senators” for their respective states. Their objective? To develop a National AI Policy.

To prepare, students are completing 8 to 10 hours of pre-work, analyzing artifacts that the Institute will synthesize into a working template. This sets the stage for an authentic legislative process, mirroring the deliberation of the Senate.

The resulting student-developed policy will be distributed nationally to all superintendents by AASA. This model holds the potential to be replicated in local districts, providing a framework for involving parents and stakeholders in governing this powerful technology.

“I think the most amazing part about this event is I honestly have no idea what the kids are going to come up with,” says Jeff Riley, Executive Director of Day of AI and former Massachusetts Commissioner of Education. “Kids are digital natives. This is an opportunity where their voices can really be heard as leaders in the field.”

Whitaker was thrilled to receive the invitation, viewing it as a natural extension of CSD’s core mission. “The stars really aligned,” she says. “This experience will take the work we do at CSD to elevate student voice to the next level."

By giving students a seat at the table both locally and nationally, CSD ensures that AI policy isn't just something done to students, but with them. This leadership experience ensures that the district’s Portrait of a Graduate includes the ability to navigate and actively govern the technologies of the future.

As Whitaker says, the ultimate goal is both simple and profound: “To ensure you’re graduating students who are ready to engage in the world knowing that AI is there.”

Edcafe AI is a rare tool in the world of AI as it, unlike many others out there, is specifically designed to support classes through the entire teaching cycle.

Built for educators, this AI tool works to offer support across the full teaching process, from lesson planning to assessments -- all in one easy-to-access platform.

The idea here is to offer a more time-efficient way to work, without jumping around between tools. From building a lesson from scratch -- differentiating for mixed abilities -- to tracking student progress, this tool should streamline that process for ease.

This guide aims to lay out all you need to know about Edcafe AI for use in your class.

What is Edcafe AI?

Edcafe AI is a digital AI-powered assistant for teachers to help across planning, teaching, and assessing lessons. By combining multiple tools, this offers a single platform for teachers to create, organize, deliver, and evaluate learning.

This allows teachers to plan curriculum-aligned lessons and then create classroom-ready learning materials. Assigning these to students, and assessing as well as tracking progress, is also built into the platform.

That all should mean teachers only need to open this one tool to get everything done, saving on time and effort -- so they can focus on more valuable teacher-student time.

How does Edcafe AI work?

Edcafe AI combines the power of artificial intelligence with the materials already being used by teachers. Current and sourced materials can be uploaded for use, all being enhanced to work with the AI ecosystem.

What that actually looks like is a teacher uploads a document, slides, or curriculum content. The AI then generates lessons, quizzes, or activities that can be shared using links or QR codes. Students can then work with these materials as teachers track engagement and performance in real time.

Usefully, there are feedback loops that allow teachers to see how students are responding, so as to refine future lessons based on student ability.

Since students don't need accounts, this is all super simple to setup and share so students can begin working with the materials right away.

What are the best Edcafe AI features?

Edcafe AI uses artificial intelligence to make the creation of materials fast and easy from full lesson plans, slides, and quizzes to reading activities.

Chatbots can also be created, with custom inputs, which are then able to assist students as they work through topics. They then can get personalized feedback and help as they work, without it costing teachers on time.

Those quizzes that can be generated are also able to be auto-graded, saving even more teacher time. This can be a quick way to assess student ability with any learning gaps easy to identify and work on improving.

Since Google Drive can be used with this system, it makes for an easy way not only to access and enhance current materials, but to generate and share more across the school.

Everything is end-to-end here, so teachers can work in one platform to get everything done. This is not only helpful for saving on time, but can also be far less mentally straining.

How much does Edcafe AI cost?

Edcafe AI offers a tiered pricing model, including a free option. The plans break down like this:

Free plan: Limited AI generations and core tools.

Pro plan: From around $7.99/month (annual billing) with expanded features.

Premium plan: From around $14.99/month with unlimited usage and school-level tools.

Custom options are also available for schools and districts, including onboarding and integration support.

Edcafe AI best tips and tricks

Start with your own materials
Upload existing lesson content to get more accurate, curriculum-aligned outputs.

Use chatbots for differentiation
Set up different chatbots for varying ability levels or topics to personalize support.

Generate then refine
Treat AI outputs as a starting point and tweak for tone, accuracy, and class context.

A recent Sports Illustrated article described how some high school football players are using artificial intelligence to fabricate highlight plays in recruiting videos. Generative tools can now insert touchdowns, tackles, or explosive runs that never actually occurred, creating a misleading picture of an athlete’s performance.

At first glance, the story reads like another example of technology being used to gain an unfair advantage. But in many cases, these videos are not as convincing as students might think.

Experienced recruiters, coaches, and even knowledgeable fans can often tell when something is off. The flow feels wrong. The movement does not match the context of the game. What appears impressive at first glance can quickly raise questions.

The real problem is that students may not realize how easily these videos can be spotted. They may overestimate how convincing these tools are and underestimate how quickly credibility can be lost. When that happens, it is not the technology that carries the consequences, but the student–and their integrity.

The Stigma of Consequences

To be fair, authenticity challenges in sports are not new. Highlight reels have always been curated, with athletes and coaches selecting the best moments, angles, and plays. Recruiting videos have never been neutral.

At the same time, experienced evaluators bring a trained eye to the process. They understand the rhythm of a game, the timing of movement, and the mechanics behind each play.

Manipulated clips raise immediate suspicion as it is not that hard for trained adults, and even knowledgeable fans, to recognize one. And when one is exposed, students are the ones who carry the consequences. The loss of trust, the damage to reputation, and the stigma of being caught can follow them well beyond a single video.

In that sense, integrity and reputation are not secondary concerns, but the foundation of the entire process.

Discouraging Shortcuts

AI did not create this challenge, but it has made it easier to act on poor judgment. What once required significant effort can now be done quickly and with little technical skill. The tools are more accessible, and the temptation is greater.

And that makes the response even more important.

Since I am not a recruiting expert, my suggestions here may be overly simple or not fully informed. Still, it seems reasonable that technology could help protect integrity. For example, recruiting platforms or video services might use AI tools to detect manipulated footage or verify highlights against full game recordings uploaded by schools. If AI can generate fabricated highlights, it can also flag these or confirm authentic ones. Even a basic verification layer could serve as a deterrent. Technology alone, however, is unlikely to solve the issue.

Coaches and athletic programs play a critical role in shaping how athletes understand the purpose of highlight reels. AI can be used to analyze performance, break down game film, and identify areas for growth, which support student development. Fabricating plays does none of that, so helping athletes understand that distinction is part of the work.

The cultural layer also cannot be ignored. Highlight videos now live on social media, where attention can become its own reward system. Likes, shares, and visibility can create pressure to produce more impressive content. When attention becomes the incentive, exaggeration can follow.

Coaches and programs can push against that pressure by reinforcing that highlights are meant to reflect real performance, not manufactured moments. Clear expectations and consistent messaging can discourage shortcuts before any take hold.

Opening Doors

Integrity in high school sports is not enforced by rules alone, but is shaped by culture, expectations, and the adults who guide young athletes. When those signals are clear, athletes are more likely to understand that how they represent their performance matters as much as the performance itself.

AI did not create the integrity challenge in high school sports, but has made the consequences more immediate and visible. The real question going forward now becomes if student athletes will understand how easily manipulation can be detected and what is at stake when it is.

In high school sports, performance may open doors. Integrity and reputation determine whether those doors stay open.

Google is attempting to position Gemini as a leader for AI tutoring when it comes to standardized test prep for the SAT and other tests. Earlier this year, they announced a partnership with the Princeton Review to allow practice tests followed by instant feedback from Gemini that can tell a student where they excelled and highlight areas they might need to study more.

The feature builds on Gemini’s learn mode. Previously called “Guided Learning,” this is a tutoring mode built on a collection of AI models called LearnLM that were specifically designed with teaching in mind and built with principles of education science.

Here’s a closer look at Gemini’s tutoring and test prep abilities.

Using Gemini To Prepare for Tests

For SAT prep, Gemini can generate full practice tests for free. All you need to do is enter a prompt such as, “I want to take a practice SAT test,” and Gemini will generate a test on it. This test will utilize resources from the Princeton Review, a test-prep company that draws on real SAT questions to generate practice tests.

In my experience, this feature works just as advertised. It took me only minutes to generate a quick practice test, and I also had the option of having Gemini create sample SAT questions. It quizzed me and offered encouragement when I got questions right and great explanations when I got questions wrong. This is a tool that seems like it would be helpful for students preparing for this test.

This type of access to high-quality practice tests and tutoring is a great way to increase access to better test preparation methods for people from all economic backgrounds as Gemini is available for free. I also suspect Gemini’s AI coaching could be helpful even for those who get one-on-one tutoring.

Google Wants Gemini To Be A Tutoring Leader

This recent foray into test prep is part of Google’s larger effort to make Gemini the go-to AI for tutoring. So far they seem successful in these efforts to me.

Gemini in learn mode has outperformed other similar tools in certain studies. For instance, researchers compared Google Gemini's and ChatGPT's effectiveness for enhancing the academic writing skills of English learners in two Thai universities. The study found “both tools effectively improved linguistic accuracy and essay structure,” however, “Gemini was found to work better than ChatGPT in multimodal feedback and source integration.”

This tracks with my own experience using both tools. While any AI tool has its limitations and subtle inaccuracies can be a big problem for test prep, Gemini in learn mode is the best AI tutor with which I have interacted. And Gemini continues to improve its tutoring features. In addition to the SAT test, Google recently announced Gemini can provide vetted tests for the Joint Entrance Examination, a standardized engineering entrance exam used in India. It also seems likely Google will expand Gemini’s data set to include other standardized tests going forward.

Overall, I’m once again really impressed with Gemini’s tutoring abilities. My graduate school options were somewhat limited because I was so afraid of taking the GRE that any program that required its completion was ruled out. It’s possible had a tool such as Gemini been available when I was younger, I would have felt differently, though I am very happy with the graduate program in which I ended up attending.

As regular Tech & Learning readers know, I have mixed feelings about AI. But I’m impressed by this feature. Using Gemini to prepare for standardized tests feels as if it could have been a game-changer for me personally and may indeed be a game-changer for many students today.

You can watch the video above or download/listen below.

Andrew Marcinek remembers the moment social media stopped feeling like a community. It was around 2015 and 2016, when he was finishing a stint at the U.S. Department of Education and coming back into schools. He was also headlining Tech & Learning events. I’ve been lucky enough to keep in touch with Andrew since then. He continues to keep a keen eye on this incredibly complicated and harrowing landscape.

His personal reckonings from that time became the foundation for his book Teaching Digital Kindness, and it's the lens through which he's watching the current wave of lawsuits against Meta and other social media companies. For Marcinek, these cases aren't a surprise — they're a long-overdue reckoning. The comparison to Big Tobacco, he argues, is apt: years of denial, a product engineered to be addictive, and mounting evidence of harm to the most vulnerable users.

The burden of dealing with that harm, he notes, has fallen almost entirely on schools — institutions that were themselves manipulated by the same algorithms and given no roadmap for what to do.

Marcinek's prescription isn't a phone ban or a screen purge. It's intentionality — building technology programs in schools with deliberate purpose, communicating transparently with parents about how and when devices are used, and treating the current AI moment as a second chance to get the foundational work right.

"Schools need to lean in and have conversations around AI," he says. "If you're not, you are doing your school community a disservice — sending kids out into the world without knowledge of probably one of the most powerful technologies in anyone's lifetime."

He points to employer signals as the clearest argument for AI literacy: companies, including Microsoft and Anthropic, are explicitly looking for workers who understand how to use AI alongside strong soft skills. Schools that treat AI primarily as a cheating threat, Marcinek contends, are asking the wrong question — and risking the same institutional failure that defined the social media era.

A second book, tentatively titled Untangled, is in progress — it focuses on how individuals and institutions can find their way out of the algorithmic knots of the last two decades. I’m looking forward to it.

Artificial intelligence has rapidly shifted the instructional landscape. Tools that can generate explanations, draft essays, and summarize complex topics are now readily available to students. This accessibility has led some to question whether deep instructor content knowledge still holds the same importance. The answer is an unequivocal yes.

AI amplifies access to data, but it does not replace the need for human expertise to validate, structure, interpret, assess, and model disciplinary knowledge. In many ways, the instructor’s role becomes more dependent on strong content knowledge because they serve as the critical mediator between AI-generated information and meaningful learning. Many generative AI tools even articulate it at the end of a response, such as what Google’s Gemini recently told me: "AI may provide inaccurate information, so double-check its responses.” It is not a piece of window dressing; it is a sincere and serious warning.

For instance, in a recently published text about using AI in the classroom, I found an example in which AI was comparing the concerns around excessive civilian deaths in Gaza in 2024 with those at the Battle of Gettysburg. The example made it through the editorial process, but it lacks historical analysis. Someone with a strong historical knowledge base would know that, though Gettysburg remains the bloodiest battle in the Civil War and the Civil War was the bloodiest war in American history, only a single civilian death has been documented. Many historians articulate that the number is most likely underreported, but Jennie Wade remains the only documented civilian death in a three-day battle that caused 50,000 military casualties. The concerns about Gaza included that the military and civilians were often interspersed, and most of the fighters were not regularly uniformed troops; therefore, opposing troops were often unsure of their enemy. This was not an issue at Gettysburg, a battle between two regularly uniformed and disciplined armies, making the AI comparison factually inaccurate.

Information Itself Does Not Equal Understanding

One of the most persistent misconceptions in AI-rich environments is the assumption that access to data equates to understanding. AI tools can produce fluent, coherent responses across nearly any subject area. However, these outputs are fundamentally probabilistic, often reflecting patterns in data rather than verified truth or disciplinary consensus.

Students often lack the background knowledge needed to distinguish between accurate synthesis and plausible error. Without a strong knowledge base, they are vulnerable to accepting AI output at face value, which creates a new instructional imperative. Teachers must understand their content deeply and be able to identify subtle inaccuracies, omissions, and misrepresentations in AI-generated material.

Content knowledge becomes the filter through which AI is evaluated. Without that filter, both instructors and students risk confusing fluency and responsiveness with accuracy.

From Content Delivery to Knowledge Architecture

As with the advent of previous technologies, it seems that some believe AI will reduce the need for instructors to function as primary sources of information delivery, not unlike how a speaker at the opening of Heidelberg University in 1500 anticipated teachers being replaced by books. Obviously, that has not come to pass, and instead, new sources of information delivery have reframed the role of the teacher.

Effective instruction now centers on the design of coherent knowledge systems. Strong instructors sequence content intentionally, emphasizing foundational concepts and building toward more complex applications. This structured progression is not something AI inherently provides. AI delivers fragments, and instructors need to construct the architecture for the overall arc of any content. Fully understanding vertical and horizontal curricular alignment (how courses fit into the bigger picture) is essential to effective instructional development.

Students require guidance in understanding what matters within a discipline, how ideas connect, and why certain knowledge is foundational. These decisions depend on deep familiarity with the field. In this sense, content knowledge is about knowing how knowledge within a given subject is organized.

Interpretation and Context in an AI Environment

AI responses often lack context and may provide correct information but fail to situate it within appropriate contexts. This limitation becomes especially significant in disciplines in which interpretation is central.

Instructors with strong content knowledge bring nuance to the learning process. They can explain competing perspectives, highlight areas of debate, and connect ideas to real-world applications. They can also help students understand when AI outputs oversimplify complex issues. This interpretive role is essential and ensures that learning moves beyond surface-level engagement and into deeper disciplinary understanding.

Assessment in the Age of AI

Assessment practices are under increasing pressure in AI-rich classrooms. Traditional assignments that rely on recall or basic synthesis are easily completed with AI assistance.

This reality demands a more sophisticated design. Instructors must create tasks that require application, judgment, and contextual understanding, which are areas in which content knowledge is indispensable. Designing meaningful assessments requires a clear sense of what constitutes deep understanding within the subject area.

Evaluating student work also becomes more complex as instructors must distinguish between superficial correctness and genuine comprehension. This distinction is only possible with a strong grasp of the content. AI raises the stakes of assessment since it makes shallow tasks obsolete and elevates the importance of disciplinary expertise.

Modeling Disciplinary Thinking

Students do not just need information but also to learn how experts think. AI can generate answers, except it does not model the reasoning processes behind those answers in a way that supports learning.

Instructors play a critical role in making their thinking visible by demonstrating how to approach problems, analyze evidence, and construct arguments within a discipline.

This modeling extends to the use of AI itself. Students must learn how to question AI outputs, refine prompts, and evaluate responses–all skills are grounded in content knowledge. Without this, students cannot effectively engage in this process.

Equity and Access to Knowledge

The implications of content knowledge extend beyond instructional effectiveness and are deeply connected to issues of equity. Students who enter classrooms with a strong background knowledge are better positioned to use AI tools effectively. They can ask better questions, recognize stronger answers, and build on existing understanding. Students without that foundation are more likely to rely on AI passively.

This dynamic risks widening existing gaps. Knowledge-rich instruction becomes a critical lever for equity by ensuring that all students develop the background knowledge necessary to engage meaningfully with AI and other information sources.

Instructors’ expertise allows them to identify essential knowledge and ensure it is taught systematically and coherently.

The Reinforced Role of the Instructor Moving Forward

The net effect of AI integration is a reinforcement of a teacher’s role, as it increases the premium on content knowledge because it raises the stakes of interpretation, validation, and meaningful use. Instructors are now becoming validators of accuracy, designers of knowledge systems, interpreters of complexity, and architects of assessment. The presence of generative AI makes these roles more visible and more critical.

Educational conversations about AI often focus on tools, policies, and academic integrity, but none of this should overshadow a more fundamental truth: Strong instruction in an AI-rich classroom depends on strong content knowledge.

Investing in teacher expertise is essential. Professional learning, curriculum design, and instructional planning must all prioritize deep disciplinary understanding. Technology will continue to evolve, and as it does, strong teachers remain key to effective instruction.

A recent analysis from Stanford University looked at more than 150,000 prompts from more than 4,400 K-12 teachers interacting with AI. The research provides a rare look at how teachers are actually using AI in practice, and most of these uses are fairly positive.

Teachers, the research suggests, are generally using the tool to help them build curriculum content and as a sounding board, all with an eye toward better helping students.

The research was conducted in partnership with SchoolAI, an AI-powered educational platform. Though it was based on data from Fall 2024, it still provides a compelling snapshot of how, where, and when many students are turning to AI.

Cynthia Chiong, Principal Research Scientist at SchoolAI, recently shared some of the highlights of the research with me.

1. Ways Teachers Are Using AI: More Than 40% of Teacher Prompts Focus On Curriculum

Chiong says this is a good sign, as it suggests teachers are using the tool to explain learning and not just automate tasks. In separate conversations with educators, she has found that many of these efforts are geared toward personalizing learning for students.

"For example, these three students here really want to focus on the sun,” she says. "While these other students are struggling with something else or are interested in another topic."

Chiong says that for a lot of teachers, being able to meet all these students where they are is the biggest draw of using AI.

2. AI As A Sounding Board

The research found that about 1 in 7 prompts involved a teacher reflecting or processing without making a specific request. This includes sharing a frustration, thinking through a challenge, or working something out.

Chiong says that these types of conversations and others teachers had with AI show that it’s an active process for educators. They might start with a question such as, “'How might I assess this math project that we've been working on?' but then they immediately go to a next thread of, ‘Okay, now, how do I put this back into a lesson to teach?’” she says.

Chiong adds, “So it's not like a one-and-done, but they're really thinking through the trajectory of learning for their students, and that was really cool to see.”

3. More Than 50% of Prompts Ask It To Generate Materials Such as Lesson Plans, Assessments, or Feedback

Teachers obviously find AI helpful when it comes to generating class material, but Chiong suggests that the data also shows teachers are not just plugging in a prompt and settling for the first output.

“The teachers are really driving the content,” she says. “It's not that the AI is like, ‘Oh, here is a lesson, this is how you should teach it.' But rather a teacher is saying, ‘Hey, I really want to teach something about evolution or whatever,' and that idea of how to teach it is still coming from that teacher entering that prompt.”

4. Half of Teacher AI conversations Include Fewer Than 10 Prompts

Teachers, by and large, are keeping their conversations with AI short, sweet, and to the point.

“They have a good sense of what they want to teach, and they have the curriculum already,” Chiong says. “So it's more a matter of translating what they already do into this new AI platform, and so it shouldn't take all that much time, and it is meant as more of a time saver and not a time taker.”