Schools Chose AI Detection Over AI Fluency. Now Your Kid Is on Your Clock.

More than half of America’s K-12 teachers cannot tell you what their school’s policy on AI is. Not because they weren’t paying attention. Because there isn’t one. Fifty-two percent say their school has offered no guidance, or they have no idea what the guidance even is.

Meanwhile sixty-two percent of those same teachers are already using AI to do their own jobs.

Read those two numbers together and you have the whole story. The people standing in front of the class have quietly decided this technology is real. The institution around them has decided nothing at all. That gap is where your kid is sitting right now.

The numbers come from a fresh NPR/Ipsos poll, fielded this spring across 545 K-12 teachers. It is the most honest snapshot we have of what is actually happening in American schools, and it is not the story you’ve been told. The story you’ve been told is that teachers are scared of AI and want it kept out. The data says something far more interesting. The teachers aren’t the problem. They’ve already crossed over. Sixty-nine percent of the ones using it say it makes them more productive. Seventy-two percent say it improved the quality of their teaching materials. They are using it to build lesson plans, write quizzes, draft the parent emails, catch up on the paperwork that was eating their Sundays.

So this isn’t a story about resistance. It’s a story about a vacuum.

What the system chose by choosing nothing

Here is what the system decided, in the absence of an actual plan. It decided AI is a cheating problem to be policed, not a skill to be taught.

Schools are building an AI police force instead of an AI class.

Look at where the energy is going. Sixty-four percent of teachers want their district to hand them free AI-detection software. Thirty-nine percent have already started requiring more assignments by hand. Another thirty-nine percent are pushing more work back into the classroom where they can watch. The instinct, everywhere, is to catch it.

Now look at where the energy is not going. Only thirty-three percent of schools have any formal policy on how students should use AI. Only twenty-one percent provide students with AI tools at all. Just thirteen percent have redesigned a single assignment to assume AI is part of the work, the way the real world already does.

The split is the entire problem. We are spending on enforcement and starving instruction. We are teaching kids to hide that they used the tool instead of teaching them to use it well.

And the enforcement doesn’t even work. Among teachers who’ve tried AI-detection software, only thirty-six percent find it reliable. Forty-two percent have never touched it. So the move on the table, the thing two-thirds of teachers are asking for, is to spend district money on a tool that doesn’t function, to police a skill the same district refuses to teach. That’s not a strategy. That’s a flinch.

I want to be fair to the teachers here, because their fear is not dumb. Fifty-five percent say AI is mostly a shortcut students use to dodge real work. Fifty-four percent say it makes critical thinking harder. They are right that a kid who hands his thinking to a machine learns nothing. He just gets a cleaner-looking version of having learned nothing. That failure mode is real, and any honest person building with these tools has watched it happen. But you don’t solve it by banning the calculator. You solve it by teaching the kid what the calculator is for, and what it can’t do, and how to know when it’s lying. That’s the part nobody is funding.

I’ve watched this exact movie before

I have some standing to call this, because I’ve sat through the earlier screening.

When I was building Buzznet and the early sites that became Buzzmedia, we were reaching more than a hundred million people a month before the phrase “creator economy” existed. The schools of that era were busy installing web filters. The internet was contraband. It was the thing you snuck onto in the library, the distraction to be blocked, the threat to be contained. The institutions treated the most important medium of the next twenty years as a discipline issue.

And the notion that a kid could make a living on it, build an audience, tell a story, get paid for both, was a punchline. The grown-ups in charge laughed at it. They were stuck in a past that was already ending, and they couldn’t see the next state of tech, or of story, coming straight at them.

The kids who ignored them built that next twenty years.

And here’s the tell: I’m not even the one saying this one is big. The teachers are. Seventy-four percent of them say AI in education has bigger implications than computers or the internet did. These are the people watching it land on actual children every day, and three out of four of them are telling you this is the largest thing they’ve seen. Seventy-eight percent say teaching responsible AI use should be part of the curriculum.

They want to teach it. The building won’t let them, because the building hasn’t decided what “it” is.

What AI fluency actually looks like

Let me make this concrete, because “AI literacy” has already been worn smooth by people who’ve never built anything with it.

I work with these tools every day at CO/AI. I don’t mean I prompt a chatbot for a haiku. I mean I build with them: drafting, researching, automating the production work that used to take a team. And the single most valuable skill in that whole process is not “knowing the prompts.” It’s judgment. It’s knowing the moment the machine is confidently wrong, which it is constantly, and catching it before it ships. It’s making the tool show its work so you can check the work. It’s treating the output as a draft from a brilliant, tireless intern with no common sense and zero accountability. Useful as hell, trusted with nothing.

That’s fluency. Not “uses AI a lot.” Uses AI and knows when not to.

A school that gets this would teach verification as a core subject, the way it once taught kids to check a source in a library. It would assign work that assumes the tool exists and grades the thinking the kid did on top of it. It would teach the limits as hard as the capabilities. Almost no school in this survey is doing that. They’re buying detectors.

Which makes this your job now

So here’s the part where it lands on your desk, because you knew it was coming.

Your kid’s school is, on average, two to four years away from a real AI curriculum. The survey makes that timeline plain. In the gap, fluency doesn’t stop developing. It just goes private. The kids who get it will get it at home, from a parent who pays for the good tools and sits next to them while they learn what the machine is good for and where it lies.

Which means the fluency gap is about to become a class gap. The kids of operators, people who already run businesses on this stuff, will walk into the next economy fluent. Everyone else will walk in policed. That’s the real cost buried in these numbers, and no detection contract is going to fix it.

You can’t wait for the school. So here’s the playbook, free, because the most useful thing I own is the inside mechanic and you should have it.

Treat it like teaching them to drive. You sit in the passenger seat. Put the real tool in front of them, not a locked-down toy version, and let them use it on actual homework, out loud, with you there. Make them show the work and show what the AI gave them, then make them find the part the AI got wrong. There’s always a part. Finding it is the whole lesson. Teach them the one sentence that matters more than any prompt: the machine sounds most confident exactly when it’s making things up. Make verifying a reflex, not a chore. Do this twenty minutes at a time and your kid will be more AI-literate than ninety percent of the workforce inside a year, and you’ll have taught the judgment the school is too frozen to touch.

More than half of America’s teachers can’t tell you their school’s AI policy. The ones who’ve used it think it’s bigger than the internet. Both of those things are true at once, and the kid in the middle of them is yours.

Nobody is coming to close that gap on a useful timeline.

You’re the curriculum now.

---

Anthony Batt is co-founder of CO/AI and co-host of the Future-Proof podcast. He co-founded Buzzmedia and the VR studio Wevr, and builds with agentic AI tools daily. Survey data: NPR/Ipsos, “Education and AI,” 545 U.S. K-12 teachers, April 27 to May 5, 2026.