What to Actually Tell Students About Using AI (A Policy You Can Use Tomorrow)

Ask ten teachers what their AI policy is and you'll get ten different answers — usually somewhere between "banned, full stop" and "I haven't really thought about it." Neither holds up. Students are using AI whether there's a policy or not, and a vague rule is no rule at all.

Here's a framework you can adapt and explain to a class in under two minutes, plus the exact language to use.

Start with the principle, not the rule

Don't lead with "you can't use AI for X." Lead with what you're actually trying to protect: their thinking. Try this line, word for word if you like:

"AI is a tool, like a calculator or spellcheck. The question isn't whether you're allowed to use it — it's whether using it here helps you learn or lets you skip the learning. I'll tell you which is which for each task."

That single sentence does most of the work. It frames AI as task-dependent rather than universally good or bad, which is closer to how you'll actually want to apply it across different assignments.

Use a traffic light system per task

Rather than one blanket policy, label each task green, yellow, or red and say so explicitly when you set it.

Green — AI encouraged. Brainstorming topics, checking grammar, explaining a concept a different way, generating practice questions to self-test. Tell students: "Use AI here. I want you to."

Yellow — AI allowed with disclosure. Drafting an outline, getting feedback on a draft, summarising research before they write their own analysis. Require a one-line note: "I used [tool] to [what]." This is the category most teachers skip, and it's the most useful one — it teaches disclosure as a habit instead of treating AI use as something to hide.

Red — AI off. In-class essays, assessments measuring a specific skill, anything where the mark is meant to certify they personally can do the thing. Be specific about why: "This task is checking whether you can structure an argument yourself. That's the skill being graded, so this one's red."

Write the three categories on the wall or in your task instructions. It takes five minutes to set up and removes the ambiguity that causes most disputes later.

Give them the prompt for the grey area

Students will still ask "is this cheating?" about something genuinely ambiguous. Give them a question to ask themselves instead of asking you:

"If I had to explain how I got this answer, in my own words, with the AI window closed — could I?"

If yes, the AI use was a tool. If no, it replaced the learning. This reframes the issue around understanding rather than catching them out, which is also just a more honest way to think about it.

Model disclosure yourself

If you use AI to draft a parent email, write a rubric, or plan a lesson, say so out loud occasionally: "I used AI to draft this and then edited it." Students copy what they see normalised, not what they're told in a one-off talk at the start of term. A teacher who treats disclosure as unremarkable gets students who do the same.

The one thing to avoid

Don't make the policy about detection — "I can tell when it's AI-written" is usually false, occasionally true, and always a worse foundation than a policy built on trust and clear categories. Detection-first policies turn into an arms race. Categories-first policies turn into a habit.

TeachSmarterAI helps New Zealand teachers use AI tools effectively. New posts every week.