The short answer
- A free, copy-paste AI policy built on the three-tier model, so students always know what’s allowed on each assignment.
- Adapt it to your subject and voice — and label each assignment AI-free, AI-assisted, or AI-encouraged.
- Keep the disclosure rule. Requiring students to note any AI help is the most protective habit you can teach.
- Pair it with the parent letter and AI-resistant assignments.
Most AI policies fail by being vague. This one is built to be specific: it uses the three-tier model so students always know, per assignment, what’s allowed. Copy it, adapt it to your subject and voice, and share it on day one.
AI in this class
AI tools (like ChatGPT) are powerful, and you’ll use them in your future. My goal is to help you use them to learn, not to skip the learning. So here’s how AI works in our class.
The one rule: use AI to understand your work — never to replace it. Using AI to explain a concept, check your reasoning, or make practice is fine. Turning in AI’s writing or answers as your own is not, and is treated as academic dishonesty.
Every assignment is labelled with one of these:
- 🚫 AI-FREE — No AI use. These assignments measure what you can do. Using AI here is a serious integrity violation.
- ⚙️ AI-ASSISTED — AI is allowed for specific purposes (explaining, checking, feedback on your own draft). You must disclose how you used it. This covers most homework.
- ✅ AI-ENCOURAGED — You’re expected to use AI and reflect on how it helped. These assignments are partly about learning to use AI well.
Always disclose. Whenever you use AI on an AI-assisted or AI-encouraged task, add a short note: “I used [tool] to [do what].” Being open about it is exactly the habit I want you to build — and it protects you.
Why these rules? School is where you build skills you’ll need for the rest of your life. AI can help you build them or help you fake them, and only one of those is still standing on test day. I’d rather you learn the thing.
Questions? Ask me. If you’re ever unsure whether AI is allowed on something, ask before you use it, not after.
How to make it yours
Adapt the wording and the tiers to your subject. What counts as “AI-assisted” in a maths class (checking a method) differs from an essay class (feedback on a draft), and a policy that ignores that gets ignored right back. Rewrite it in your voice — a policy that reads like it came off the internet gets treated like it came off the internet.
Two things to keep whatever you change:
- The disclosure requirement. It’s the habit that generalizes to college and work, and it’s what protects a student if a detector wrongly flags them. If you keep one rule, keep this.
- The “why.” Students follow rules they understand. The short explanation of why the rules exist earns compliance that “because I said so” never will.
Put it into a system
A policy works best when the assignments and grading back it up. Pair this with AI-resistant assignment design (so following the policy is the path of least resistance), rubrics that reward thinking (so fluent AI prose can’t score well anyway), and the parent letter (so families reinforce the same expectations at home).
Together they turn a policy from a paragraph students skim into a system they actually operate inside.
Frequently asked questions
What should a classroom AI policy say?
It should state, in plain language, which uses of AI are allowed and which aren’t (ideally per assignment), how students must disclose any AI help, what happens if the policy is broken, and why the rules exist. The clearest policies label each assignment as AI-free, AI-assisted with disclosure, or AI-encouraged, so students never have to guess whether a tool is permitted.
How do I use the three-tier AI model in practice?
Mark every assignment with one of three labels: AI-free (no AI, for assessing individual skill), AI-assisted (AI allowed for specific disclosed purposes), or AI-encouraged (students expected to use and reflect on AI). Put the label on the assignment itself. This removes the ambiguity that causes most AI policy problems, because expectations are explicit every time.