AI Is Already in the Classroom. Most Teachers Say No One Gave Them the Rules.

The number from this week’s NPR/Ipsos poll that grabbed the headlines is 54: the share of K-12 teachers who say artificial intelligence is making it harder for students to learn to think critically. It’s a striking figure, and it’s also the least actionable one in the survey. Teachers’ fear that a new tool dulls young minds is as old as the tool. Calculators were going to ruin arithmetic. The internet was going to end deep reading. Some of that worry ages well and some of it doesn’t, and a single poll of perceptions can’t tell you which.

The number that should actually move a school board is quieter. In the same poll, based on 545 teachers surveyed April 27 to May 5 with a margin of error of about five points, only around a third of teachers said their school has formal guidelines for how students may use AI. Nearly eight in ten said schools should teach responsible AI use. Set those two findings next to each other and you have the real story: the people closest to the students are asking for a rulebook, and most of them say they don’t have one. That gap, not the 54%, is the part adults in charge can fix this year.

The tool is already inside the building

Whatever districts decide, they’re deciding late. The poll found that AI is not a hypothetical knocking at the schoolhouse door; it’s already at the desk. Roughly two-thirds of teachers said they use it to create classroom materials, about half to plan or write lessons, and others for administrative work and parent communication. On the student side, more than half of teachers said AI isn’t being used in their classrooms yet, but about two in five said students are using it in class at least once a week.

That’s diffusion, not adoption-by-plan. The technology entered through a thousand individual decisions, a teacher saving an hour on a worksheet here, a student opening a chatbot on a phone there, long before any committee voted on a policy. Nearly three in four teachers told the pollsters AI will have bigger implications for education than the internet or computers did. They are watching the most consequential classroom shift of their careers happen without a framework around it, and a majority say they can’t find one where they work.

This is the same dynamic that keeps tripping institutions up everywhere AI lands. The capability arrives first and cheaply; the governance, accountability, and norms arrive late and contested. It’s a pattern visible in everything from the fights over who’s responsible when an AI product harms someone to the overheated claims about AI erasing entry-level jobs. Schools are not special here. They’re just slower, and the stakes are children.

The rules may exist on paper and still be missing in the room

There’s a complication worth being honest about, because it cuts against the easy version of this argument. By some measures, schools have been busy. More than half the states have issued AI guidance for K-12 districts, and the count has kept climbing past 30 states. Ohio went furthest, becoming the first state to require every district to adopt a formal AI policy by July 1, 2026. And in a separate survey of district technology leaders, the Consortium for School Networking’s State of EdTech 2026 report found that nearly 80% say they have AI guidelines in place, up sharply from a year earlier.

So which is it: a third, or four-fifths? Both, and the distance between them is the most useful thing in any of this data. District tech leaders say the policies exist. Most teachers say they can’t find them. A guideline that lives in a PDF on a central-office server, written by people who don’t teach, is not a guideline a ninth-grade English teacher can lean on at 7:45 a.m. when a student turns in an essay that reads like a machine wrote it. The failure isn’t only absence. It’s policy that doesn’t reach the classroom in a form anyone can use.

What a real answer looks like

The fix here is unglamorous, which is exactly why it gets deferred. It is not a moratorium, and it is not a districtwide ban that every teenager defeats by lunchtime. It’s three concrete things.

First, grade-appropriate rules that say plainly what counts as acceptable AI use and what counts as cheating, written in language a 14-year-old and a substitute teacher can both follow. Second, actual instruction: treat AI literacy the way schools eventually treated internet literacy, as a skill to be taught rather than a temptation to be policed. Eight in ten teachers already say this is the job; they’re right. Third, distribution: get the policy out of the binder and into orientation, syllabi, and the daily expectations students actually see.

None of that requires resolving the hard philosophical question of whether AI is making kids worse thinkers. It just requires deciding that a powerful tool already in your students’ hands deserves the same boring institutional seriousness you’d give a chemistry lab or a phone policy.

Why this can’t wait

The honest caveat is that the alarming 54% is a measure of what teachers believe, not of what AI is provably doing to young brains. We should hold that finding loosely; the research on AI and cognition is young, and educators who feel besieged may read more harm into a new tool than the evidence yet supports. Skepticism cuts both ways.

There’s a human cost to the drift, too, and the same survey caught it. Seven in ten teachers said they think the public’s perception of their profession has gotten worse. A workforce that already feels under suspicion is now being handed a disruptive, cheating-adjacent technology and told, in effect, to sort it out on its own. Asking teachers to police a tool no one has given them rules for is a fast way to deepen exactly that resentment. A clear policy isn’t only an academic-integrity document; it’s a signal to teachers that the institution has their back.

But the policy gap doesn’t depend on settling the cognition debate. Even if AI turns out to be a net gift to learning — a tireless tutor, a leveler for kids without expensive help at home — students still need to be taught how to use it without outsourcing the thinking, and teachers still need backing they can point to. The cost of waiting isn’t measured in some distant test score. It’s paid now, every week, by the teacher improvising a rule on the spot because no one above her would write one down. The poll didn’t really find that teachers are afraid of AI. It found that they’re being asked to manage it alone. That’s the finding a district can do something about before the next school year starts.