Teaching students to work with AI

AI policy for your syllabus

Most AI syllabus statements fail the same way: they tell students what they can’t do, and leave them guessing about everything else. Here are four policies you can copy, and the one sentence that makes any of them work.

A syllabus policy on AI has one job. When a student is sitting with the assignment open at 11pm and a chatbot in the next tab, the policy has to answer the question they’re actually asking: is this allowed, and how would anyone know?

“Use of AI is prohibited” doesn’t answer it. Neither does “use AI responsibly.” The first is unenforceable and the second is meaningless. What follows are four policies that do answer it, at different levels of permission, plus the disclosure line that makes any of them workable.

Four policies you can copy

1. Open use, with disclosure

Best for skills courses, project work, and anything where the deliverable matters more than the drafting.

This works because the disclosure is small enough to actually comply with and specific enough to be informative. “What you changed” is the load-bearing phrase - a student who changed nothing has to either say so or lie, and lying is a much clearer violation than the fuzzy offence of “using AI too much.”

2. Open use on process, closed on the final artifact

Best for writing-intensive courses where you want students to develop their own voice but not to suffer unnecessarily.

The five-minute conversation test is worth keeping even if you never hold the conversation. It gives students a concrete standard, and it gives you a fair procedure if you ever need one.

3. Assignment by assignment

Best for courses with varied assessment types, and honestly the most accurate reflection of how most instructors actually feel.

This is more work to maintain and much easier for students to follow. The default-to-Assisted clause saves you from having to label everything perfectly.

4. Restricted, with the reason stated

Best for foundational courses where the skill being built is the one the tool replaces.

If you restrict, say why. A ban with a reason is a teaching moment. A ban without one reads as fear, and students discount it accordingly.

The sentence that makes any of them work

Whichever policy you pick, add this:

Two things happen when you include it. Students actually ask - which tells you far more about how AI is being used in your course than any detection tool will. And it removes the excuse that the policy was unclear, which is the defence you will otherwise hear in every integrity conversation you have this term.

Four mistakes worth avoiding

• Relying on a detector. AI detectors have false positive rates high enough to accuse innocent students, and they disproportionately flag non-native English writers. Don’t build a policy on top of a tool you can’t defend in an appeal. More on why detection fails.
• Writing a policy you can’t observe. If you cannot tell whether a rule was followed, it isn’t a rule, it’s a wish. Restrict where you can see the work.
• Treating all AI use as one thing. “Asked it to explain a concept” and “submitted its output as my essay” are not the same act. A policy that can’t distinguish them will be ignored for the first and useless against the second.
• Leaving it out of the assignment sheet. Nobody re-reads the syllabus in week nine. Put the label on the assignment itself, every time.

Common questions

Do I have to allow AI at all?

No. But you should decide deliberately rather than by default, and you should be honest with yourself about which of your assessments can actually enforce a restriction. Most take-home written work can’t.

What if my institution has its own policy?

It takes precedence, and most institutional policies explicitly delegate the specifics to the instructor. Your syllabus statement usually sits underneath it, adding the detail the institutional policy leaves open.

Should students cite AI like a source?

Formal citation is awkward, because the output isn’t stable or retrievable. A short disclosure note is more useful and much more likely to be completed honestly. If your discipline’s style guide has a format, use it - but ask for the disclosure either way.

How do I handle a student who clearly used AI when it wasn’t allowed?

Have the conversation, not the accusation. Ask them to walk you through their process and defend their choices. A student who did the work can do this easily. A student who didn’t will usually tell you so within a few minutes, and you’ll have a far more defensible record than a detector score.