How to Write an AI Policy for Your Course (With Example Wording)

Why does each course need its own AI statement?

Because students cannot assume one rule applies everywhere. The same student might be encouraged to use AI in marketing and banned from it in a writing seminar, in the same week.

The University of Toronto's instructor FAQ on generative AI notes that students get different instructions from course to course, and asks instructors to state clearly whether and how AI is allowed on each assessment. U of T's October 2026 sample syllabus language also advises keeping conditions simple, since students are juggling AI rules from several courses at once.

Some institutions now require a statement. UBC's Senate Policy V-130, for example, requires every course syllabus to say how generative AI use is or is not permitted in student work. Check your own institution for local rules.

What should a course AI policy include?

A usable policy answers the questions students actually ask.

• Your stance and why. Tie it to a learning outcome, such as building your own analytical writing.
• What counts as AI. Chatbots, plus AI features in Word, Google Docs and Grammarly.
• Where it applies. Which assignments, and which stages: research, outlining, drafting, editing.
• How to disclose. A template to paste, plus how to cite any AI output they quote.
• Course materials. Whether students may upload your slides, readings or rubrics to AI tools. U of T strongly recommends a separate statement on this.
• Questions and follow-up. Who to ask, and a note that you may ask any student to explain their work.

Three sample AI policy statements you can adapt

Canadian guidance tends to use three stances: McMaster's course outline help offers prohibited, some use permitted and unrestricted options, and Waterloo's course outline suggestions include text for permitting, conditionally permitting or forbidding AI. The statements below were written for this guide; replace the brackets and adapt.

1. AI not permitted on assessed work

Best for courses where the assessed skill is the one AI would replace, such as foundational writing.

2. AI permitted for specific uses, with disclosure

Best for most courses: AI helps where it supports learning, and students still do the core work alone.

3. AI encouraged, with disclosure and verification

Best for upper-year or applied courses where using AI well is itself a goal.

How do you set AI rules for each assignment?

The syllabus sets your default; each brief sets the specifics. UBC's guidance notes that rules may need to differ by assessment, or even by stage, such as allowing AI for an essay outline but not on the exam.

These rows are examples, not rules for your discipline. Copy the relevant line into each brief.

A disclosure statement students can paste

Require one template on every assignment, with a no-AI option, so disclosure is routine rather than a confession.

Disclosure does not replace citation: quoted AI output still needs a citation in your course's style.

Should AI-detection tools enforce your policy?

No. Detector scores are not reliable enough to support an integrity decision, and errors fall hardest on students writing in an additional language.

• OpenAI withdrew its own detector. OpenAI's AI text classifier, released in January 2023, caught 26% of AI-written text in OpenAI's own tests and labelled human writing as AI 9% of the time. OpenAI said it should not be a primary decision-making tool, and on July 20, 2023, withdrew it "due to its low rate of accuracy."
• Detectors misread non-native writing. In a 2023 paper in Patterns, Liang and colleagues at Stanford tested seven widely used detectors on 91 TOEFL essays by non-native English writers and 88 US eighth-grade essays. The US essays were classified accurately, but more than half of the TOEFL essays were labelled AI-generated (average false-positive rate 61.3%). Enriching the TOEFL essays' vocabulary with ChatGPT cut false positives sharply, so detectors are also easy to game.
• Some universities do not support them. U of T does not support AI-detection software on student work, citing unreliability, false flags on human writing and the privacy concerns of uploading student work.

Instead, talk with the student, look at their notes and drafts, and ask them to explain their reasoning. At U of T, meeting with the student is the first step when misuse is suspected.

How do you design assessments that show learning?

Make the thinking visible and add moments where students show understanding in person. U of T's FAQ concedes that traditional assignments may no longer reliably assess learning.

1. Collect process evidence. Mark a proposal, outline and draft, not just the final file. Waterloo's sample statement asks students to keep notes, drafts and prompts, ready to submit.
2. Add an in-class component. A short in-person task extending the take-home work, as U of T suggests, shows whose understanding it is.
3. Hold brief oral check-ins. Five minutes per student, or a rotating sample, explaining one choice. Announce it in the syllabus so it feels routine, not accusatory.
4. Anchor tasks in your course. Require a specific class discussion, data set or lab result.
5. Weight the rubric toward reasoning. Give more marks for argument and use of course concepts, fewer for surface polish.
6. Test your assignment with AI first. Run your brief through an institution-approved chatbot and mark the result. If it scores well, redesign or add a process step.

Showing students that output beside strong work makes expectations concrete. Never paste student work.

How should you communicate your AI policy?

Say it in three places: the first class, your learning management system (LMS) and every assignment brief. U of T advises telling students clearly, both verbally and in assignment instructions.

1. First class. Explain the policy and its learning reason, show one acceptable and one unacceptable use, and invite questions.
2. LMS. Pin the policy, the disclosure template and a running FAQ. UBC recommends telling students how to ask about specific tools or scenarios.
3. Each assignment brief. Add the one-line AI rule and the disclosure requirement.
4. Your TAs. Give them the same rules in writing.
5. Next offering. Review the policy each time the course runs; tools change quickly.

Where CMPUS fits

You can run a course edition of CMPUS built on your syllabus, readings, rubric and AI-use policy, so students get AI help inside your course's rules. Students work with Classmates such as Jordan, who gives a practice grade with criterion-by-criterion feedback against your rubric, and Quinn, who builds practice tests from course notes. Classmates are not designed to hand back finished assignments.

CMPUS is not an AI-detection or surveillance product. As with any outside tool, check your institution's rules on approved tools and course materials first. Details are on the universities page.

Instructor resources from Canadian universities

• University of Toronto: Generative AI in the Classroom FAQs, with downloadable sample syllabus language.
• UBC: Generative Artificial Intelligence Syllabus Language, with course-wide, assignment-specific and disclosure examples.
• University of Waterloo: Course Outline Suggestions for Generative AI, with sample text for three stances.

Questions people ask

Sources

1. Generative Artificial Intelligence in the Classroom: FAQs · University of Toronto, Office of the Vice-Provost, Teaching & Learning
2. U of T Sample Syllabus Language Regarding Use of Generative AI in Courses (October 2026) · University of Toronto
3. Generative Artificial Intelligence Syllabus Language · University of British Columbia, Centre for Teaching, Learning and Technology
4. UW Course Outline Suggestions for Generative Artificial Intelligence · University of Waterloo
5. Why are there three options for Generative AI policies? · McMaster University, Course Outline Help
6. New AI classifier for indicating AI-written text · OpenAI
7. GPT detectors are biased against non-native English writers (Liang et al., 2023) · Patterns (Cell Press)

CMPUS guides are general advice, not your course's rules. Your syllabus and your instructor decide what is allowed. Spotted something out of date? Tell us.