Free for Teachers, Banned for Students: The Split

There's a scene playing out in American school districts that looks, from a distance, like institutional whiplash. A district blocks ChatGPT on the student network, cites academic integrity, tells kids the tool is off-limits. Then the same district's teachers sign up for ChatGPT for Teachers — free, sanctioned, unlimited — and start drafting lesson plans with the exact model their students just got banned from using. Baltimore, Los Angeles, New York City, and Seattle all restricted student-facing ChatGPT at various points; none of those restrictions touch the teacher offer, because the teacher offer was never a student product. The obvious reaction is that somebody is being a hypocrite. The less obvious, and more accurate, reaction is that these are two different questions that happen to share a brand name.

The bans were never about the technology

When a district blocks ChatGPT for students, the stated reason is almost always academic integrity — kids submitting AI-written essays as their own, offloading the thinking the assignment was supposed to produce. That's a real problem, and it's worth naming precisely: it is a problem about assessment, not about AI. The essay is a proxy. We ask students to write essays because writing one forces the reasoning we actually care about. A tool that produces the essay without the reasoning breaks the proxy. The ban is an attempt to protect the proxy while the institution figures out what to do.

The Week's coverage of the cheating debate captures the bind districts are in: they can't easily detect AI writing, they can't un-invent the tool, and they can't pretend students won't reach for it. Blocking it on the school network is a blunt instrument, and everyone involved knows it's blunt. It buys time. It signals a norm. It does not solve the underlying question of how you assess learning when the artifact of learning can be manufactured on demand.

What matters for the split is this: none of that reasoning has anything to do with a teacher using the same model to build a differentiated reading passage. The academic-integrity objection is entirely about the student doing the student's own work. Point the tool at the adult who designs the work instead, and the objection evaporates — there's no assignment to cheat on, no proxy to protect, no grade at stake.

The teacher offer is a different bet entirely

OpenAI's ChatGPT for Teachers help documentation is explicit about scope: the free workspace is for verified U.S. K-12 teachers, school staff, leaders, and district administrators. It is not a student login. Students in most of these districts do not get free ChatGPT access through this offer — the program was designed around the adult in the building, not the child at the desk.

That design choice is what makes the two positions coherent instead of contradictory. A district can hold both of these beliefs at once without flinching:

• Students writing essays with ChatGPT undermines the thing the essay was measuring.
• Teachers drafting lesson plans with ChatGPT reclaims hours they'd otherwise spend on prep.

The tool is identical in both sentences. What changed is who's holding it and what they're being asked to produce — and that difference is the entire argument.

The teacher isn't being assessed. Nobody grades a lesson plan for originality. When a teacher uses the model to generate three tiers of a worksheet, the output is input to their professional judgment — they edit it, reject it, adapt it. When a student uses the model to generate an essay that gets graded, the output is the assessed artifact. Same model, opposite relationship to the work.

Where the split gets genuinely uncomfortable

None of this makes the arrangement perfectly clean, and it would be dishonest to pretend otherwise. There are at least two places the split rubs.

The first is the optics problem. It is genuinely awkward to tell a sixteen-year-old "you may not use this tool, it's a shortcut that cheats you out of learning" while their teacher openly uses it to shortcut lesson prep. The teacher has a ready answer — my prep isn't the thing being assessed, your essay is — but that answer requires a level of nuance about the purpose of assessment that most policies never bother to explain to students. When you don't explain the reasoning, the rule looks like "adults get the good toy and you don't," which is exactly the kind of message that breeds cynicism.

The second is the modeling problem. Students are going to enter a workforce where fluency with these tools is assumed. A blanket student ban, paired with enthusiastic teacher adoption, risks teaching kids that AI is something you use in secret once you're old enough, rather than something you learn to use well and disclose openly. That's a bad lesson, and it's the opposite of what most of these districts say they want. The honest version of the teacher-only offer isn't "students can't handle this" — it's "we haven't yet built the assessments that survive students having it." Those are very different admissions, and only one of them is about the students.

This tension isn't unique to OpenAI's launch; it runs through the whole debate about who these deals actually serve, which is the subject of the conflict-of-interest question in AI training elsewhere in this series. And the skepticism it generates on the ground — teachers wondering whether "free for us" is a gift or a wedge — is exactly what surfaced when Oregon teachers called AI deals a "grift".

So is it coherent policy or not?

Here's the resolution: it's coherent as far as it goes, and it doesn't go far enough. The teacher offer and the student ban are not contradictory, because they answer different questions. "Should a graded student essay be allowed to be AI-generated?" and "Should a teacher's lesson prep be allowed to be AI-assisted?" have different right answers, and a district that says no to the first and yes to the second is being consistent, not hypocritical.

What that district is not doing — what almost none of them are doing — is answering the question that actually matters: what happens to students? A teacher-only rollout is a stable position for exactly as long as you refuse to look at the student side. The moment you do, the ban reveals itself as a placeholder. It's not a philosophy of how kids should relate to AI; it's a pause button held down while the assessment problem stays unsolved. The foundational facts of what ChatGPT for Teachers actually includes make the teacher side generous and clear. The student side is the part still written in pencil.

The split, in other words, isn't the problem. The split is a reasonable interim answer to a real distinction. The problem is treating an interim answer as if it were a finished policy — courting teachers with a polished, free, sanctioned product while leaving the harder half of the question, the half about the kids, exactly where it was before the launch. Free for teachers is easy. Figuring out what students deserve is the work nobody's shipped yet.

Part 71 of 100 in the ChatGPT for Teachers series. Previously: Oregon Teachers Are Calling AI Deals a "Grift". Next: The Conflict-of-Interest Question in AI Training. Browse more builder insights or explore AI skills for education at aiskill.market.