How can teachers set clear AI rules for the new school year?

This is the first full school year that begins with ChatGPT freely available to students from day one. Many schools still have no rule about it. OpenAI says as much in the educator FAQ it published on August 31: many districts and universities have not yet updated their academic honesty policies for generative AI, and each institution will settle the question on its own timeline. Until your school decides, the clearest rules are often the ones a teacher writes for one class.

Clear rules do two jobs. They tell students what is allowed before they start, so nobody has to guess. And they give you something fair to point to when work does not match the rule.

Start from the assignment, not the tool

A single rule for the whole year ("no AI" or "AI is fine") rarely survives contact with real assignments. A spelling quiz, a lab report and a debate preparation need different answers. A simpler approach is to label each task with one of three levels:

Level	What it means	Example task
No AI	The work shows what the student can do alone	In-class essay, reading check, quiz
AI for named steps	AI may help with steps you list, such as brainstorming or checking grammar; the student writes the final work	Research project, lab report
AI open	Any use is fine if it is disclosed and checked	Debate practice, revision quizzes

UNESCO's quick start guide on ChatGPT in higher education, published in April, notes that some institutions responded by banning the tool while others changed how they assess, moving toward in-class or non-written work. The three levels let you do both, task by task.

Don't build the rules on detectors

It is tempting to rely on software that claims to spot AI writing. OpenAI's educator FAQ answers the question "Do AI detectors work?" with "In short, no." It explains that no detector, including one OpenAI built and tested, has proven able to reliably separate AI-written from human-written text. When OpenAI tried, its detector marked passages of Shakespeare, and even the Declaration of Independence, as machine-written. The FAQ adds that detectors may unfairly flag students writing in a second language or in a very formulaic style, and that small edits defeat them anyway. Asking ChatGPT "did you write this?" is no better: it has no way of knowing, so its answer is made up.

A rule that depends on a detector will catch some innocent students and miss others. Rules built on visible work are fairer and easier to explain.

Make the work visible

The approach OpenAI suggests is to ask students to share their ChatGPT conversations, using ChatGPT's share links, alongside the finished work. That gives you three things: you can see how the student thought through the task, you can grade the quality of their questions and their checking, and the student knows from the start that use will be seen. Drafts, notes, outlines and short conferences ("walk me through your second paragraph") do the same job for any assignment.

OpenAI's guide for teachers, published the same day, includes a useful example from a high school computer science teacher in Chennai who tells students that ChatGPT's answers may not always be accurate and should be confirmed through other sources. That habit belongs in the rules too: if students use AI, they are responsible for checking what it says.

Mind the age rules and privacy

ChatGPT's terms require users to be at least 13, and students between 13 and 18 need a parent's or guardian's permission. For younger classes, that settles the question: any AI use happens on the teacher's screen, with the teacher typing. For all ages, students should never paste in personal details about themselves or classmates.

What should a class AI policy include?

A good class policy fits on one page and answers the questions students will actually ask: When can I use it? How do I show that I did? What happens if I don't? Here is a worked example for a grade 9 English class. Adjust the levels and examples to your subject.

Our class rules for AI tools (2023–24) 1. Each assignment has a level. It is printed at the top: No AI, AI for named steps, or AI open. If it is missing, ask before you start. 2. No AI means none. That includes rewording tools and chatbots. 3. For named steps, use AI only for those steps. The assignment lists them, for example "brainstorm topics" or "check grammar after your final draft." 4. Show your use. Hand in the share link to your ChatGPT conversation, or a copy of it, with your work. Add one sentence: what you asked, and what you kept or changed. 5. You are responsible for every fact. AI can be wrong and can invent sources. Check anything you keep against a book, an article or another reliable source. 6. Keep your drafts. I may ask you to explain or extend your work in a short conversation. 7. Protect privacy. Never type in names, grades or personal details about yourself or anyone else. 8. Age rule. ChatGPT requires users to be 13 or older, with a parent's or guardian's permission if you are under 18. 9. If the rules are broken, we talk first, and you redo the work under the right level. Repeated cases follow the school's academic honesty policy.

Try it: Take your first three assignments of the year and give each one a level. For any task marked "AI for named steps," write the exact steps in one line, then read the line as a student would and fix anything that could be read two ways.

Share the page with families in the first week and go through it with students using one real assignment, so they can see what a shared link and a one-sentence note look like. Revisit it after the first term; your school may publish its own policy by then, and your class rules can simply point to it.