We had Justin Reich and Jesse Dukes as guests on the Silver Lining for Learning webinar/podcast to discuss their new guidebook, A Guide to AI in Schools: Perspectives for the Perplexed. The resource, based on over 120 interviews with teachers and students, offers a snapshot of what educators and students in the US are actually thinking and doing with this new technology.

Note: You can watch the video on the Silver Lining for Learning site or follow us on your favorite podcast app.

As I read the report and reflected on our conversation, I was struck by a curious tension: how conventional most AI use in schools remains, and how unconventional the concerns teachers express about it are. That gap is worth exploring.

Larry Cuban and others have documented for decades how schools absorb new technologies into existing practices rather than being transformed by them. The “grammar of schooling,” the deep structures of how schools organize time, space, subjects, and assessment, proves remarkably durable. Just as grammar shapes what we can say in a language, these institutional structures shape what’s possible in schools, channeling new tools into familiar patterns. The Reich/Dukes interviews confirm this pattern. In fact, Jesse Dukes, reflecting on all the interviews, said that he couldn’t think of a single educator they had spoken to “who has really transformed what they’re doing with AI.”

For the most part teachers were generating lesson plan drafts, creating rubrics, differentiating reading materials, writing emails to parents—essentially productivity tweaks and workflow assists. Students, meanwhile, use AI to get unstuck on homework or understand difficult texts. Pretty standard stuff.

Yet despite the conventional uses, teachers express a series of existential concerns about this technology. Some are professional and pedagogical. Many teachers, especially English teachers, worry about what AI means for the core of their work. Sara Falls, an English teacher in San Francisco, puts it sharply in the guidebook: “Writing is thinking. And it’s hard. Now more than ever students need to work to cultivate the patience, the openness, and the insight to do this without AI.” The worry isn’t just about cheating. It’s about whether the processes through which students develop, struggling with ideas, finding their voice, learning to think by writing, get short-circuited.

Other concerns extend beyond pedagogy into broader societal and ethical territory. What is AI doing to the environment? To democracy? To the information ecosystem? To young people’s relationship with truth? One environmental science teacher refused to use AI precisely because of its energy demands. Others worry about students forming parasocial relationships with chatbots, about deepfakes, about a generation that can’t distinguish AI-generated text from human writing.

These broader concerns, I think, are the more significant ones, precisely because they point beyond what classroom-focused research can capture.

I’ve argued elsewhere that while we conduct careful studies comparing learning conditions, we miss the larger story. The real cognitive and cultural transformation isn’t happening in classrooms. It’s happening through algorithmic systems optimized for engagement, in private conversations with AI companions, in the slow reshaping of young people’s relationships to truth and institutional authority.

The Reich/Dukes guidebook illustrates this pattern. It documents AI use inside the grammar of schooling: lesson plans, rubrics, academic integrity, differentiation. These matter. But the existential concerns teachers express point toward something bigger and deeper: the transformation happening outside school walls. When teachers worry about democracy, environment, and the collapse of shared truth, they’re intuiting something important: the real educational significance of AI may not be about classroom integration at all. Film didn’t transform classrooms, but it transformed the world within which classrooms exist. The same may be true for AI.

To be fair, I’m not entirely sure what the next steps are, or how we maintain academic rigor while seeking to understand and tackle these gnarly socio-technical issues. What is clear to me, however, is that we need new paradigms and new approaches, ones that recognize that the most important questions about AI and education won’t be answered by studying classrooms alone.