What Students Want Teachers to Know About AI

They Want the Same Thing They've Wanted For More Than Two Years. Help.

Back in December I sat down with a handful of HS students on a couple different occasions to talk about AI - their thoughts, attitudes, and descriptions of how students were using it in school. There were lots of observations - most unsurprising - but an overall confirmation that student AI usage had become a mainstream part of their academic lives. With the 2025 - 2026 school year more than half over and as we headed into a much needed spring break, I had an opportunity to sit down with about a dozen HS students in our AI club - 9th, 10th, and 11th graders - to revisit the conversation about AI and school.

One asked about the recent issue between Anthropic and the Pentagon.¹ Another was willing to share how he uses AI as a study aid. And several talked about how AI was impacting the corporate world. On the whole, the recent conversation made something even clearer - the AI gap between teachers and students continues to grow despite more and more schools trying to catch up.

I know this is one group of kids at one school - members of an AI club, not exactly a representative sample, though one that notably included several girls whose voices have often been absent from this conversation.² But I suspect the dynamic is recognizable to teachers well beyond our building.

These kids were thoughtful, specific, and honest. They had ideas, critiques, and an underlying message - please help us understand how AI fits or doesn’t fit into school because we are using it ALL. THE. TIME.

How They’re Actually Using It

I opened the meeting with a simple question: how are you and your peers using AI?

One student shared some screenshots with me of several ways he uses AI as a learning tool. Using Claude’s presentation feature, I stitched them together into a short deck, which he walked us through. He showed us interactive math diagrams with adjustable parameters, AI-generated quizzes with built-in explanations, flashcard sets with shuffle modes, and gamified study tools - all generated from very simple prompts.

Most educators (me included) wouldn’t characterize this as AI misuse. This student has figured out that AI can function as a personalized tutor, study partner, and test designer. When I first started writing about AI a little more than a year ago, few if any of these features were available. The AI issue has rapidly moved beyond mere text generation.

Another student described using AI for product research and hobby work:

“It’s really good at taking two products and comparing and contrasting them in a way that Google can’t. If I give it two things I’m looking at, it’ll give me pros and cons.”

When I asked whether he trusted the output, he didn’t hesitate:

“For any research I’m doing, I ask it to give me the main points of what I’m researching with citations, and then I go through those citations to see what’s actually critical.”

That was encouraging though I suspect it’s not the norm.

The “Neanderthal” Problem

One student offered an opinion with typical teenage bluntness:

I think for a lot of people my age, their understanding of AI is pretty Neanderthal. I think right now a lot of people just see it as a shortcut versus something you integrate within your daily life to improve your efficiency. There’s just so many more uses that people overlook - they see it as a very two-dimensional thing.

This obviously complicates the narrative that students are simply using AI to cheat. These students - granted, again, these are kids in an AI club - were drawing a clear line between those who use AI as a shortcut and those who use it as a tool - and they were frustrated that teachers, and even their own peers, couldn’t always see the difference.

Another student connected this to what awaits them in the future:

I think a lot of people skip over the fact that AI is being used way more in the workforce. I’ve heard a lot of people who work in technology and business - they’re starting to use AI models because it’s just a way more efficient way to get stuff done. I think that’s going to be really interesting to see the landscape of the next twenty years.

I told her I wasn’t sure it would take twenty. More like five.

What They Want Teachers to Know

I tried to shift the discussion from description to prescription. I asked the group directly: if you were a teacher right now, how would you handle AI in your classes?

One student referenced something one teacher had done - create a custom chatbot designed around specific assignment guidelines that students could use within defined parameters:

One thing I really liked is [when a teacher] made ... a chatbot thing that was specifically designed to follow the guidelines [they] set up. We could go into it, and it was limited. I think that's a good way to do it, because telling kids just the guidelines - like, [another teacher's] guideline for AI is just none, zero - I think that's very hard for students, because then it makes every single thing they do with AI feel very bad. And they’re still going to do it.” [emphasis added].

This is probably least surprising but also most difficult to swallow. Students will use AI whether we forbid it or not. That’s just the reality.

When I asked her to confirm - students are using AI regardless of the policy? - she didn’t hesitate:

“Yeah, definitely … I think teachers should be aware that kids are going to use AI. And I think I would want to create my own space that has guidelines and make sure all the students are in there, under my supervision, learning how to use it effectively. Because I think learning how to use it effectively is better than trying to make kids not use it.”

Another student made the case against prohibition even more directly:

“I think criminalizing AI, especially in the classroom, does way more harm than allowing it in moderation. If AI is just outright banned, it gains this association of being negative, like it’s cheating, like it’s bad. Whereas if students are taught to use AI in moderation - because they’re going to use it, they’re going to use it regardless - this way it teaches them to be conscious and aware of their AI usage and learn how to use it effectively.”

One more offered a practical recommendation:

“If I was a teacher, in any other class like English or creative arts, I would actually teach how to use AI to make a study guide for large blocks of information. It’s beneficial for memorization and studying key terms. But if it’s an in-class writing assignment, a skills check - you shouldn’t use AI just to get the work done faster. If it’s studying, it’s one thing. If it’s writing whole sentences for you, whole paragraphs, that’s another thing.”

These kids are proposing precisely what many commentators are advocating - context-dependent policies that distinguish between AI as a study aid versus AI as a substitute for thinking.

Pushing It Underground

A repeated observation popped up about the difference between AIs with guardrails and typical models found online. This is the problem Marc Watkins tackled in a recent post - what we might call the “free version” problem. A student who had experience with our AI wrapper remarked:

“I used Flint on my own time, and I have to say, compared to other models, it was incredibly frustrating because it’s designed the way teachers want you to use it. It wouldn’t give me direct answers - it would be like, ‘What do you think?’ It wasn’t that helpful in my opinion.”

Which is exactly what Flint and similar paid AI wrappers are supposed to do - push the student to think rather than hand them answers. Institutions are pouring money into these products. But the notion that kids will default to AIs deliberately designed to create “friction” when ample free versions exist that do not is more than unrealistic. It’s fantasy. When I asked how many had used or were even aware of the “study mode” drop-down selections baked into all the frontier models, they just stared at me.

Another student was even more direct about what happens when schools take a punitive approach:

“I think a big issue with how schools are dealing with AI right now is that so many kids are getting in trouble for it, and I don’t think it necessarily discourages the use of it. If anything, it just pushes it underground even more. I think making it into a learning movement - learning from mistakes and understanding the line of what’s okay and what’s not - rather than just punishing would be more effective.”

Obviously, these are teenagers. They didn’t catch the contradictions (introduce guardrails but we don’t want guardrails!) nor do they see the larger issues through the same lens as experienced teachers, almost all of whom learned their academic skills in a world where generative AI didn’t exist. These kids don’t know that world. They are trying to adapt in real time while many teachers are still assigning the same kinds of work that can now be done faster and better by AI than by high school students. That’s not their fault.

How Will We Close the Gap?

When students consistently said that teachers should be teaching them how to use AI effectively, I shared something I’ve observed in my conversations with educators across the country: many teachers don’t know how to use AI effectively. Quite a few are openly critical, many more are ambivalent at best, and those few who do feel comfortable enough to teach with it often find themselves navigating colleagues and administrators who are dealing with a hundred other priorities.

The result is that student use of AI is probably more experimental and varied than what most teachers are doing. Only a small fraction are particularly fluent with it. Many have serious reservations - and I understand why.

One student responded:

“I think a lot of teachers see AI as a bad thing because they don’t really know how to use it themselves. They only think of students plugging in a prompt and it just spitting out an answer. But I think if teachers develop this AI literacy - I think AI might see a lot more healthy use in classrooms than it’s seeing right now.”

He was getting at something I’ve written about before - the idea that teacher engagement with AI, even imperfect engagement, might shift student behavior more than any policy could:

“I think by teachers showing students how to navigate the ropes of AI, it creates an environment where students might respect the boundaries a bit more, yet still use AI in a way that’s beneficial to them but not overly harmful.”

I don’t know if he’s right about that, and I suspect many teachers would disagree. When I asked how many of their teachers had ever designed an AI-integrated lesson or even demonstrated what they consider appropriate AI use, the answer was almost none.

What I Showed Them

Towards the end of our club meeting, I shared some of the things I’ve been building (which I documented in my previous post) - an interactive pre-Civil War module to support a discussion-based activity, a Constitutional Convention simulation where students adopt delegate personas, and a data visualization of American war casualties over the centuries. I also showed them Claude’s newest visualization features and a history simulator built by Ben Breen, a college professor, to give them a sense of what other educators were experimenting with.

Nearly everything I build with AI stays within my subject area, and I frequently use my own materials - primary sources, discussion questions, lesson content - to help create additional documents and assignments. But the best way to understand the limits of AI is to test it on something you don’t know well.

When I suggested on the spot that I demo something in STEM (not my area of expertise!), a student suggested an interactive graphic of Bohr’s model.

I was honest:

“When I use AI, I tend to use it in subject matters where I know a lot, so I can get a sense of whether it’s way off. And for the most part, it’s really pretty good. So I’m going to extrapolate that this is still decent. But I don’t know - if I showed this to a physics teacher, he might be like, ‘Oh, this is totally wrong,’ and I would have no idea.”

The point wasn't to impress them. It was to model what using AI as a design tool looks like - and to be transparent about its limitations. I can build rich, interactive history modules because I provide the content and generally know when something is wrong. Expertise has to come first. AI is most powerful when the human in the loop actually knows what they're doing.

What’s Changed Since December

In my December conversations, students described, in part, a world of clandestine workarounds - split screens, retyping AI output by hand, and switching platforms when one gets banned. The vibe was deception and an acknowledgment of a cat and mouse game. How can I use it without getting caught?

This conversation was different. Not because these students aren’t using AI to get through some of their schoolwork. I’m not so naive as to think none of them do. But in the discussion on the day before break, they were willing to go further than that. They had more opinions about what responsible AI use looks like and specific proposals for what teachers should do differently, probably because I asked. They had critiques of existing tools and policies that were articulate and nuanced.

The fundamental dynamic still hasn’t changed with students still navigating AI largely on their own and minimal guidance from teachers. The gap between what students know about AI and what most educators can offer is also real.

One student observation from the December conversation still holds:

“A lot of adults have this idea of AI and all these things about AI, and they think they’re so right about it. So I think they should be more open to changing their views on how another demographic - like students - may use AI, and avoid the stigma around it.”

The Underlying Ask

What struck me most about this conversation wasn’t any single quote. It was the consistency of the comments. Across different students, different grade levels, and different levels of AI sophistication, they were all saying versions of the same thing: they want help understanding this new technology and how to use it for learning. Telling us not to use it and just punishing us when we do is not a long-term solution. Give us a space where we can be honest about what we’re doing without fear that honesty becomes a way to indict us.

That’s not an unfair ask.

Whether most schools can deliver on that right now - with the training gaps, philosophical objections, institutional inertia, and the sheer pace of technological change - is anyone’s guess, but I know more schools and more teachers are trying. Students will speak up if given the chance, and their voices still need to be heard.

The question is how many of us are listening.

---

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

1. I spent a good chunk of time on this since I am well-versed in the issues surrounding the government contract, Anthropic's red lines on surveillance and autonomous weapons, the administration's pushback, and the troubling questions about military uses of AI. That's a post for another day.

2. Much of the student AI commentary skews male. In this conversation, three female students were among the most articulate and engaged voices. Their perspectives helpfully added to the gender gap in the AI discussion.