K-12 Can’t Borrow Higher Ed’s AI Playbook & Kids Live in the Gap

A few weeks ago I sat in a room with four high school students and a few dozen teachers who had agreed to hear what kids actually do with AI. The students were sharp, honest, and a little exasperated. In art and business class, AI was built into the projects. In biology, it was encouraged for building study guides. In English and history, it was banned. Same building, same week, four different sets of rules, and nobody could tell them why. When I asked whether anyone had ever taught them what AI actually is, how it works, where it falls apart, they looked at each other. Not in middle school. Not since.

That is the human impact of AI on education, and it is not what most of the conversation is about.

The loudest voices on AI and learning right now come from higher education, and they have mostly settled on a reasonable position: let students use it, teach them to use it well, and redesign the work so the thinking still has to happen. I agree with every word of that for a 20-year-old. The problem is that it quietly assumes a person who already knows how to learn, who has a working model of what a source is, who can notice when a confident machine is confidently wrong. A seventh grader does not have that yet. That is not a knock on seventh graders. It is the job description of seventh grade.

So when K-12 borrows the higher-ed playbook, something gets lost in translation. “Use AI thoughtfully” becomes “use AI.” And in a classroom of 12-year-olds, “use AI” is not an instruction. It is the absence of one.

I want to give the skeptics their due here, because the teachers I work with are not confused about the stakes. At a back-to-school session this fall I asked a room of veteran teachers, many with 20 or more years in the classroom, how they felt about AI and assessment. The words that came back were “wobbly,” “scary,” and “hard to know if it’s still their thinking.” They are right. A text-only assignment can be finished by a machine in seconds.

The detectors do not work, and the probabilistic math behind these tools makes it unlikely they ever will. Revision history can be faked by a kid who types a generated essay back in by hand. And we are now a year into tools that will not just draft the assignment but fill out the form and press submit.

None of that argues for banning AI. It argues against vagueness.

Here is the pattern I see in school after school. Leadership, wanting to be balanced, lands on a stoplight: red means no AI, yellow means AI with permission, green means go ahead. It is a sensible visual. Then I watch a faculty try to define yellow and green, and the room splits.

To one teacher, green means the student may use one approved tool for one purpose across a unit. To another, green means the student can turn in AI-generated work. To a third, yellow and green are basically the same thing. These were thoughtful, experienced people, and they could not agree. If the adults cannot say what the middle means, the kids certainly cannot.

And the middle is where the danger lives. Red is easy. A clear green, with a named tool and a named purpose, is manageable. It is the vague yellow, the shrug, the “you can use it a little,” that sends a student to whatever chatbot is open on their phone at 9 p.m. with no guardrails, no teacher visibility, and no instruction on what the tool is for. The school’s content filter does not follow them home. Most major chatbots no longer even require an account. The default setting for a kid with a fuzzy instruction is an unguarded tool and a growing habit of outsourcing the part of the work that was supposed to build the brain.

This is the human impact. Not a robot replacing a teacher. A child quietly learning to skip the struggle because no adult told them which struggle was the point.

What fills the silence is not complicated, but it has to be said out loud, by a teacher, for every assignment. Name the tool, and make it one designed for kids, with boundaries built in and a window for the teacher to see what is happening. Name the use: review the content, test your understanding, connect it to where you live, get feedback on a draft. Name what is off limits: do not let it draw the diagram, write the explanation, or produce the thing you are turning in. Then ask the student, in their own words or their own voice, to reflect on how they used it and whether they kept ownership of their thinking. A 10-second verbal explanation tells me more about understanding than a page of polished text ever could.

Underneath all of it, the thing that makes any of this work is a relationship. Students who feel caught in a “gotcha” culture get sneaky. Students who are trusted and clearly guided tend to rise to it. The four kids on that panel were disappointed in classmates who cheat, not because they feared getting caught, but because it cheapened their own effort. They wanted the adults to stop assuming the worst and start explaining the rules.

K-12 does not need higher ed’s answer. It needs its own: developmental, explicit, and spoken aloud. The question is not whether kids will use AI. They already do. The question is whether anyone has told them what for.