Where Do Schools Go From Here?
Four years of conflict have exhausted both teachers and students. Is it a stalemate, or a detente?
Since school ended a few weeks ago, I have pulled back a bit from the AI conversation. I needed to get outside and away from my computer after a very busy spring. The World Cup has helped, though not as much with the getting-outside part. The rule changes have improved the viewer experience and the U.S. Men’s team is onto the knockout stage, which is a bonus.
Watching the matches is a reminder of the first time someone demoed ChatGPT for me. It was the first weekend in December 2022, the weekend the U.S. lost in the Round of 16 to the Netherlands in Qatar. I was running a debate tournament. We had a big screen up to display the pairings, and another coach pulled up ChatGPT, which had debuted a little over a week earlier. I hadn’t heard of it. I asked it to write a poem about the World Cup in the style of Shakespeare, which it did, in seconds. I was amazed. What an innocent time!
After millions of words written about AI and schools, here in the summer of 2026, I think students and teachers have arrived at something like a detente. If we intend to move forward productively in the fall, we need to stop fighting about AI long enough to ask what school is actually for.
The AI Arc (Late 2022 - Summer 2026)
Generative AI arrived out of nowhere, and no one was hit harder, day to day, than teachers and students. Amazement was immediately followed by fear - mostly of cheating. For some it generated excitement, then anger and backlash as people learned more about how these systems were trained, what they cost, and their environmental impact. A 2026 Brookings report on K-12 found that, on the current trajectory, the risks of generative AI in schools outweigh the benefits. Skepticism was present from day one - both about the technology itself and even more about predictions of AGI, mass unemployment, and the end of humanity.
Ironically, most of the apocalyptic messages come from the people closest to the technology, who continuously tell us it will eliminate jobs, surpass human intelligence, and remake society in ways they themselves describe as dystopian. The founders seem genuinely surprised by the backlash, even though the frustration so many educators and the public feel about AI is a rational response to being told the technology will upend their lives.
For students, access is now ambient. Anyone with an internet connection can pull up ChatGPT or their favorite model, which is orders of magnitude more powerful than the ones that shipped in 2023. Every assignment now carries a decision for them - how much AI to use, or whether to use it at all. Many are already dependent. But plenty remain wary, and more than a few are frustrated and angry that they were placed in this position without being asked. On that, teachers and students are often in agreement.
The AI question has both prompted and been folded into a larger conversation about phones, laptops, and the screens we have handed children going on two decades.
Schools, caught in the middle, have gone in wildly different directions. Some went all in, embedding AI into their infrastructure and their mission. Some refused and banned it outright, and many are still on their heels, yet to engage with the issue head on, churning out increasingly convoluted AI policies they have no way to enforce.
Most landed somewhere between deer-in-the-headlights and reluctant adoption, pushed along by district and university partnerships with AI companies often made under pressure to do something and not be perceived as falling behind. A few have taken thoughtful steps to confront the situation as it actually exists, resisting both the hype and the doom. And the gap between the people who use AI daily and the people who either reject it altogether or hold it at arm’s length keeps widening.
The school AI conversation sits in a larger AI media ecosystem that constantly veers from one pole to the other, making it even harder to focus on what matters amid the predictions, prophecies, and promises.
That’s the landscape I see as most schools close out the 2025-2026 school year. We’re battered, exhausted, but still standing.
Stalemate, or Detente?
My gut initially told me students and teachers have reached an AI stalemate this past academic year, like two prize-fighters circling each other in the ring, each exhausted by the flurry of punches thrown in the opening rounds. Now they’re stumbling, hoping the ref will step in and call a draw. Students cheated with near impunity for the first couple of years, and teachers fought back with invasive detection software and a return to blue book exams. Honor codes were rewritten, disciplinary outcomes were murky, and most accusations weren’t worth the time-sink to pursue.1
But this is what happens in an arms race: no advantage remains for long. Detection software was answered by humanizers built to defeat it, which were answered by better detectors, which were answered by better humanizers. Every advance invites a counter. Escalation with no decisive edge eventually grinds everyone down.
I don’t think student use of AI fell this year - the statistics tell the opposite story - but my hunch is they were more likely to use AI on the margins, where teachers could not or would not enforce strict “no AI” policies. Professors began, some enthusiastically and others reluctantly, to allow AI in limited circumstances, signaling not surrender so much as a recognition that AI had become an ever-present temptation to offload their work. Many instructors began to think more carefully about what they were asking students to do and why - a much-needed return to designing assignments around the authentic learning we claim to want. If AI leads to more curricular reevaluation that is, on balance, a net salutary effect.
What the Conflict Is Actually About
The detection war looks like a draw. It has created something closer to a detente - a condition where both sides are ready for a path forward. Students, repeatedly and consistently, have told us they want to be taught by human beings. Strip away the panic and that is what remains. The teaching relationship has always been at the heart of education.
No student ten years from now is going to tell a story about the chatbot that went off-script to tell them where it was on 9/11, or the chatbot that kept them after class to cheer them up, or the one that talked them off a ledge before a hard exam. Chatbots cannot read the mood in a room for those “teachable moments,” which are the ones students remember most. This is especially true in K-12 settings.
This is my essential observation. When that 2026 Brookings report concluded the risks of AI in schools currently outweigh the benefits, a central harm it named was the erosion of the relationship between students and teachers. The thing the research worried we were losing is the thing students keep telling me they want most. Whatever AI we let into schools has to protect that first.
This doesn’t mean AI has no role or cannot be used in ways that play to its strengths. Research is already being transformed by AI, and the ability to build bespoke tools through natural-language coding has real potential.2 But the notion that AI should become a central presence in U.S. classrooms, delivering and teaching content independently, has been undermined, especially over the past year.3
This is from my vantage point. If you are a teacher whose AI integration is really working, I’d love to hear about it. I’ve seen some of my colleagues use it creatively, and I know practitioners who are using it thoughtfully.4 But these examples feel like the exception rather than the rule, at least at the moment.
And despite what edtech vendors are selling, I have come to think most AI wrappers built for the classroom are a waste of time. For students who turn to AI to avoid their work, removing friction is the entire point; a wrapper that adds it back is a tool they won’t use. For the ones who want to use it well, they don’t want AI-lite any more than someone handed the keys to a car wants to be handed a bicycle. If we really want to teach older students how to use AI, they need access to the frontier models. For middle school and younger, classroom chatbot use is an entirely different conversation.
None of this means the detente is stable. The student who blew a deadline and is staring at a blank page at midnight on Sunday will still reach for AI, at least to get started - and no strong teacher-student relationship or coherent policy will reliably stop them. And the technology will not stay still long enough for us to get complacent. I have no idea whether the next disruption involves the proliferation of agents that move past chatbots, or ubiquitous smart glasses, or world models that create issues we haven’t even thought of. Anyone who tells you they know where all this is going is practicing a form of divination.
Student Confirmation
I’ve carried this feeling for a while this year, and recent conversations with students returning home from college confirm it. The former students I have talked with this spring keep saying versions of the same thing: most try not to use it but either can’t help themselves or have found specific tasks where it genuinely helps. More than a few have told me that they feel least guilty about using AI when they have a professor who doesn’t teach the material well in the first place.
Marcus Luther, a high school English teacher who writes really thoughtfully about teaching on his substack, The Broken Copier, asked his students again this year whether he should bring AI into his class. 81 percent said no, with only 6 percent in favor. One wrote that using AI would make them feel “way less accomplished.” His tenth graders, asked whether they wanted more or less technology in class, went around the circle saying the same word. Less. Less. Less.
While I have had success with AI tools in my advanced classes, I find myself landing closer and closer to his conviction. Using AI well in a classroom takes real expertise that most of us do not yet have, even as students keep using it everywhere else. We owe them guidance for the world they're heading into, but I'm not convinced how much time that means spending class on AI itself. That’s one of the issues I wrestle with the most. I don’t know how representative his experience is, but his student’s reactions don’t surprise me.
I imagine some of his students use AI, but my guess is not in there. He has built a culture of caring, noticing, and authenticity that makes kids want to learn how to write, how to think, and how to do the work themselves. I suspect I would have felt the same way in his class. The trouble is that not every student gets a teacher like Mr. Luther.
Angela Duckworth has described the same deep ambivalence in young people more broadly, many of them worried the technology is making them “lazy and stupid.”5 The pattern is consistent, and recent articles have named the trend: those using AI the most are often the most critical of it.6 This is the same impulse behind students booing commencement speakers this spring for invoking AI. Students are using AI more and understanding it better, but a growing number do not feel good about what it is doing to their learning.
We still do not have strong evidence that AI improves learning. A 2026 Oxford University Press study found that 72 percent of UK pupils, given the choice, declined to use AI on a school task, even as a majority used it weekly at home.7 One pupil put the distinction better than any researcher could: it “wouldn’t be what I truly think - a vague interpretation of what I’ve told a robot to expand on.” That absence does not prove AI has no place in school, but it does shift the burden onto the people promising the gains. I have found it immensely useful in my own work but have had mixed results using it with students. Talking about it with them helps. Experimenting is the only way we learn what’s effective.
More and more, I think our role in the classroom is to double down on the uniquely human connection, comfort, and guidance we provide. My ideal future is not one in which AI is absent from schools but one where students and teachers decide together what school is actually for, and only then where AI might have a place in it. I believe this will be one of the most important questions for educators over the next decade.
Why the Analogies Fail
I’ve never liked the printing-press and calculator comparisons. I wasn’t around for the printing press, and I barely remember the calculator “wars.” But AI is different in kind from other technologies that have reshaped schools. It strikes at the very heart of the learning process: it generates information instantly, answers the kinds of questions schools tend to ask more efficiently than students can, and easily undermines the very skills we want them to build. At the same time, and I can’t stress this enough, the technology is useful in the right hands and has the potential to let people do some increedible things. I reject any frame that refuses to hold both of these realities at once.
This is not an anti-AI piece. It is a pro-human one. We cannot keep bemoaning the problems AI brings without also examining the opportunities we have as educators to recommit to first principles. It means having the conversations we have spent four years avoiding - between teachers and administrators, between teachers and students, and among administrators, teachers, and parents. If we can do that, the past four years will not have been wasted. They will have set the table for the work that comes next.
A Final Indulgence
Before I published this piece, I asked ChatGPT to write a poem about the U.S. chances in this World Cup, in the style of Shakespeare - essentially the same prompt I used in 2022. Here is a snippet.
Enter Fortune, masked as VAR8
I am that glass-eyed witch whom none can love,
Who halts the shout, delays the leap, the kiss;
I pluck the eagle by the wing above
And whisper, “Was his shoulder off by this?”
The Americans are brave, and bravely young;
Their legs are thunder, yet their hearts must learn
That World Cups are not merely played, but wrung,
And half of glory is the nerve to burn.
Read it and notice your own reaction, which I think is as good a Rorschach test as any for where you stand on AI. If it still amazes you, you are probably someone who finds this technology genuinely remarkable. If all you see is a statistical engine assembling the next token, you’re probably not a fan. While the poem did not shock me the way the first one did in 2022, the VAR reference made me laugh. And a poem is now child’s play next to what these models can do.
But I think that’s the point. Effortlessly mimicking the style of a famous author about any topic imaginable is one thing; the more aware we become of what AI cannot do, the better chance we have of designing classrooms around the human qualities that make them come alive.
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.
I may be completely off base here. If you are a professor successfully using detection tools, or have a more exciting and productive experience with AI in the classroom than the one I’m describing, I’m eager to hear about it.
In his latest post, Watkins documents building online classroom tools with AI coding and includes some impressive examples. He argues the results become useful only when a human supplies the judgment, taste, and iteration. Crucially, he emphasizes that these kinds of applications can only be made thoughtfully, driven by human creativity and intention. He also observes how out of reach this is likely to be for most teachers and students at the moment.
Plenty of articles over the past few months have documented this idea, the most recent one in the Atlantic, referencing an interview earlier this year with Sal Khan who acknowledged the AI chatbot embedded in Khanmigo was a “non-event” for most students.
Mike Taubman and his colleague Scott Kern represent the kind of exception I mean, deliberately designing a course where AI is both useful and demystified. They were recently profiled on the NY Times Daily.
This is her phrasing quoted in the post from the Broken Copier.
Gallup, the Walton Family Foundation, and GSV Ventures, Voices of Gen Z: The AI Paradox - More Exposure, Less Confidence Among Gen Z (April 2026). Weekly-or-more use held at 51 percent (22 percent daily plus 29 percent weekly); excitement fell 14 points to 22 percent and anger rose 9 points to 31 percent; among daily users specifically, excitement fell 18 points and hopefulness 11 points year over year. Same finding reported by Fortune, Axios, and Education Week.
Oxford University Press, Navigating AI in Education: Pupils' Perspectives on the Role of AI in the Classroom (June 2026). The percentage figures are from the report's 3,100-respondent survey and the pupil quotation from its qualitative studies (700+ pupils across 20 schools); 72 percent of pupils with access to a generative AI tool declined to use it in the in-class writing task.
VAR stands for Video Assistant Referee, the video-review system used in soccer to check refereeing decisions on goals, penalties, and red cards. The influx of video review in major sports is another good example of how easily we get acclimated to new technology. Instant replay would have overturned many classic sporting moments over the years - Maradona’s Hand of God goal, Franco Harris’s Immaculate Reception, and Derek Jeter’s clutch “home run” in the ’95 playoffs against the Orioles. The introduction of the ABS (Automated Ball-Strike) Challenge system is just the further evolution of change. My eventual hope is that AI will be seen along these lines as well, to the point where we don’t even think about it anymore.
So much worthy of commenting on here, but I'll say that I just most appreciate the depth and scope of reflection you're modeling for others.
From the get go I argued that we needed to treat this technology as an an opportunity, not a threat, not an opportunity to invite more tech into the classroom, but as an opportunity to deeply consider what most matters when it comes to the experiences of school. I'd gotten to this place prior to ChatGPT's arrival because I'd already worked through my disillusionment with "schooling" as I evolved my pedagogy for teaching writing, but what I think you demonstrate here is that this is no longer just an individual choice. We have to look at these problems at the school and system level if we're going to do right by this generation of students.
A generous piece, as usual. I like how much time you spend on delivery, on who supplies the content, the student or the machine. On that ground, many perceive the contest is over. The machine answers the questions faster than the student. You raise this in your point about the wrappers. But when the tenth graders you report going around the circle saying less, less, less, I hear them asking for leadership. We have to articulate clearly what the work is now. The old answer, the teacher who carries and delivers the material/content into the classroom, has been taken shaken. Are we still the people who deliver the content? Or are we something else? What is our role now, as teachers? If we spend less time policing, maybe we can answer the big questions that the kids are asking (or booing for).