Teachomatic what to automate, and what not to

A Classroom AI Policy on One Page

The gap that causes most of the trouble is not students cheating. It is that more than eight in ten students report no teacher ever showed them how to use AI for schoolwork, while only about a third of teachers say they have clear guidance themselves.

So the common situation in a classroom is: everyone is using it, nobody has said what is allowed, and the first time the rule becomes explicit is when somebody is in trouble. That is not an integrity problem. It is a communication failure with an integrity problem attached.

For comparison with workplace systems that formalise time or activity data, Monitask’s page on remote employee monitoring software shows how the same type of measurement is handled outside education.

A page fixes most of it. Not a school policy — that is someone else's job and it may take a year. A page for your class, for your subject, this term.

Why one page

Because a rule nobody has read is not a rule.

For an external perspective on academic integrity, technology, and student rights, see ISTE.

Long policies are written to cover the institution. A classroom policy is written to be followed, which means a student must be able to hold all of it in their head while deciding, at 10pm, whether to paste a paragraph into something.

If it does not fit on a page, students will use the version they have inferred from your tone, and their inference will be wrong in both directions: some will assume everything is banned and quietly use it anyway; others will assume everything is fine.

Three rules to write, not one

The mistake is writing a single rule for "AI." Nobody can apply that, because the same tool doing two different things is two different questions.

Split the work instead. For each kind of task you set, decide which of three categories it falls into, and say so when you set it, not in a policy document filed somewhere.

Free. Use it however you like, no disclosure needed. Typically: getting unstuck, having something explained again, generating practice questions, checking your own understanding, translating a term.

Allowed with disclosure. Use it, and say what you used it for in a line at the end. Typically: brainstorming, structuring, feedback on a draft you wrote, help with a specific stuck point.

Not for this task. Because the task is the thing being assessed. Typically: anything where producing the words is the skill being measured.

That third category needs a reason attached, every time. "Not allowed" invites evasion. "This one is not allowed because I need to see whether you can construct the argument yourself, and I cannot see that if something else constructs it" invites compliance, because it is obviously true.

What the page should contain

Six things, and nothing else.

  1. The three categories, in your own words, with one example each from your subject.
  2. How to disclose — one sentence at the end of the work saying what you used and for what. Nothing elaborate.
  3. What happens if you disclose something that turns out not to be allowed. The answer should be: we talk, we sort it out, and it is not a disciplinary matter. If disclosure is risky, nobody will disclose.
  4. What happens if you do not disclose. Stated plainly and proportionately.
  5. What you will not do. Specifically: that you will not act on a detector score alone. Students know these tools exist and many are quietly frightened of them; saying this is one of the highest-trust sentences available to you.
  6. How the rule can change. It will. Say that you will announce it, and that changes apply forwards.

Write it with the class

Twenty minutes, once, and it is worth more than the page.

Put up three or four realistic scenarios from your subject and ask which category each belongs in. (A full lesson on how the tool works is a good session to run first.) You ask it to explain the topic again in simpler words. You ask it to write your introduction. You write an essay and ask it what is weak. You ask it to generate ten practice questions.

Two things happen. You find out what your class actually thinks is normal, which is usually surprising in both directions. And they arrive at most of the policy themselves, which means it becomes a rule they helped set rather than one imposed on them.

Where they disagree with your line, explain the reason rather than restating the rule. If you cannot construct a reason, that is information — you may have drawn the line by reflex.

The disclosure habit is the real objective

If the page does only one thing, it should establish that saying what you used is normal and safe.

That habit is more useful than any prohibition, for three reasons. It is the professional norm students are heading into. It gives you information you cannot otherwise get. And it removes the incentive to conceal, which is what turns ordinary shortcut-taking into something that looks like deception.

Which means the rule that matters most is number 3 above. Honest disclosure of something that turned out to be over the line is never punished. If you break that once, you will not get another honest disclosure from that class.

What to do about school policy

Most schools either have nothing or have something written in 2023. Your class page is not a substitute, and it is also not a rebellion — it is the implementation layer that every framework assumes exists and few provide.

Check two things: that your page does not contradict school rules, and that your school knows what you are doing. If school policy is silent, say so on your page: the school has not set a rule on this yet, so here is ours for this class. Students respect that more than a pretence of authority.

If you want an external anchor, international frameworks published from 2024 onwards — UNESCO's competency framework for teachers among them — consistently put human judgement and teacher discretion at the centre rather than tool compliance. That is a useful thing to be able to cite in a staff meeting.

What not to put on the page

A list of banned tools. It will be out of date within a term and it teaches students that the rule is about brand names rather than about what they did.

A detection threat. Beyond being unenforceable, it frames the whole thing as adversarial, and the evidence does not support the threat anyway.

Anything you will not actually do. A policy you do not enforce is worse than none, because it teaches that the stated rules are decorative. The same goes for claims about what the tools can do — the section on evidence is the place to check before repeating one.

The short version