AI Syllabus Policies

Setting an AI Policy for Your Courses

Whether and when students should use AI in your courses depends on a few important factors: the discipline/course topic, their level of expertise, the task they are undertaking, and your pedagogical goals. It is your role as the instructor to help your students avoid uses of AI that undermine their learning, and, where applicable, understand the norms for AI use in your field .

This document offers guidance for setting an AI policy that achieves your goals and matches your pedagogical aims, while also being legible and sensible for students. Your AI policy works hand in hand with your assignments, and depending on the policy you choose, you may need to make changes to your assignments to ensure that students who violate your policy cannot gain an unfair advantage over students who follow it. A policy that cannot be enforced is out of alignment with the assignments it applies to. The templates below provide starting language for policies and offer assignment-specific guidance.

Steps

Decide where AI helps and where it harms. Consider how students will best learn the material in your course. Controlled studies, for example, find that while student performance on an immediate task might improve when AI tools are producing the output, their learning gains, measured afterward without the tool, are the same (generally for tutor-type uses) or worse (generally for producing work-type uses). Generally speaking, AI is useful for some activities and detrimental for others. The deciding factor is whether the activity is designed so that the student is doing the kinds of thinking the assignment was meant to develop, rather than outsourcing that thinking to an AI tool. In field-specific courses, the norms of your field matter too: in some fields, students need to learn to use AI for certain tasks, in others the norm is not to use AI for certain tasks. Your policy should follow from this judgment, and you should share your reasoning with students.
Write the policy. We provide three starter templates below. Each has the same four-part structure: what is and is not permitted, the reason for the policy, how students disclose permitted use, and how you verify that submitted work reflects the student’s own understanding.
Communicate it. Include the policy in your syllabus, address it explicitly in class, and invite students to ask when they are unsure how to apply it. If you do not use AI to do your work, tell your students why. If you use AI to do your work, consider showing students how you use it and where you are cautious or avoid using it. Discuss with your teaching team how they will communicate the policy in section and office hours, and when and how to address potential AI misuse. Provide reminders in class or assignment-specific guidance about AI use in each assignment prompt. The assignment-specific clauses below are a starting point you can use for enacting a variable usage policy.
Adapt your assignments. Whatever policy you choose, it is important to make sure that each of your assignments will help ensure that the policy is effective. Where an assignment does not permit AI, pair it with supervised work that verifies understanding, or replace it with work that is not vulnerable to AI use. Where AI is permitted, make sure the assignment still produces evidence of what the student has learned. We give common options with the templates below, and our resource on Making Assignments AI-Resilient has detailed guidance based on assignment type.

What makes a good policy

A good AI policy is legible, such that a student understands what is permitted and prohibited on each assignment without guessing. It is sensible, meaning a student using ordinary tools can comply without violating it by accident, and it does not prohibit uses that do not interfere with the learning objectives. And it is enforceable, meaning students who violate the policy cannot gain an unfair advantage. Outside supervised settings you will rarely be able to prove a violation, so enforcement comes from the design of the course: enough of the grade rests on work that is not susceptible to AI use, supervised work, or on work that verifies understanding of submitted out-of-class work. An absolute ban stated on its own is legible but neither sensible nor enforceable.

Policy templates

Course AI policies run from permissive to restrictive, and some courses may have different policies for different assignment types. The templates below are provided for you to adapt for your course [bracketed phrases are places where course-specific customization is needed].

Restrictive policy

A restrictive policy permits little to no use of AI in producing submitted work. It fits a course or assignment where students will achieve the learning objectives by working without assistance from AI. For each major unsupervised assignment, consider whether a student who violates the policy can gain an advantage over students who follow it. If so, add or alter elements of the assignment to address that, such as in-class, supervised scaffolding or knowledge checks, or replace the assignment with one that achieves the same objectives and is not vulnerable to AI use. Our resource on Making Assignments AI-Resilient has ideas for essays, readings, problem sets, and other assignment types.

Why this policy. This course teaches [skills: writing a clear argument from evidence, reading a difficult text closely, solving problems from first principles] that you need to be able to do on your own. Using generative AI to do them for you would keep you from learning them. [Optional: In [field], the norms of scholarship and professional practice also prohibit [the uses in question].]

What is not permitted. Do not use generative AI to [produce, draft, revise, solve, translate, or summarize] any part of the work you submit, including a first draft that you then edit. This applies to stand-alone tools like ChatGPT, Gemini, and Claude [and to AI features built into software you already use, such as sentence suggestions in your word processor, AI summaries in your browser, code completion in your editor. Instructions for turning off these features are on the course site.]

What is permitted. You may use ordinary spelling and grammar checking [and web search]. You may use generative AI to [explain a concept from lecture or reading, generate practice problems], provided you then do the assigned work yourself.] No other use of generative AI is permitted in this course. If you are unsure whether a use is permitted, ask before you use it.

Verifying your understanding. For all unsupervised work for which AI-use could confer an unfair advantage, I may ask you in writing or in a conversation to explain or extend your work, or otherwise verify your understanding of what you submitted. [Optional: Where your in-class work shows much less understanding than the work you submitted, [the submitted work will be marked down / the work will be referred to the Honor Council].]

Assignment-specific policy

This type of policy fits most courses. The course-level statement sets a default, usually that AI is not permitted unless the assignment says otherwise, and assignment prompts or categories of assignments give more specific instructions about what AI use is and is not permitted. Where an assignment does not permit AI use, make sure that students who violate the policy cannot gain an advantage by doing so. Where AI use is permitted, support students in understanding which uses will serve or harm their learning, and in learning the AI skills they need to complete the work.

Why this policy. Some of the work in this course teaches skills you need to be able to do without assistance. Some of it is work where AI can productively support your learning, and some is work where AI saves time or adds capacity on tasks that are not central to course learning objectives. The policy differs by assignment so that you use AI where it supports your learning and not where it substitutes for it.

The default and the categories. Unless an assignment says otherwise, you may not use generative AI to produce the work you submit, [including through AI features built into your word processor, browser, or code editor (see our instructions for turning these features off)]. Every assignment falls in one of three categories and may include additional specific guidance about AI use.

AI not permitted [: problem sets, response papers, the midterm essay]. The restrictions above apply. You may use ordinary spelling and grammar checking and web search.
AI permitted [: the research paper, the final project]. You may use generative AI for the purposes the prompt lists.
AI expected [: the data analysis labs]. The assignment is designed to be done with generative AI. The assignment prompt explains the expected use.

You are responsible for the accuracy of anything you submit, whatever its source.

Verifying your understanding. [Assignments that do not permit AI are paired with supervised work, such as [in-class writing, discussion graded on a participation rubric, a short written or oral check after you submit].] For all unsupervised work, I may ask you in writing or in a conversation to explain or extend your work, or otherwise verify your understanding of what you submitted.

Permissive policy

A permissive policy fits two kinds of courses. In the first, scholars and professionals in the field use AI for the tasks being taught, and learning to use it well is part of the course; assignments are designed so that students meet the learning objectives while working with AI. In the second, AI use is permitted on out-of-class work because 1) it does not confer any advantage to the students who use it or interfere with learning, or 2) grading rests on supervised, device-free work, and out-of-class work with AI becomes preparation for demonstrating learning in person. In any case, it is a good idea to keep some graded work AI-free so that you and your students can assess how AI is affecting their achievement of the learning objectives. Unless you know your students already have strong AI skills, teach them how to use AI well in your discipline, and use their descriptions of how they used AI to find skill gaps and uses that substitute for learning.

Why this policy. [Choose one: Scholars and professionals in [field] now use generative AI for [the tasks this course teaches], and learning to use it well, and to know when not to, is part of what this course teaches. The assignments are designed so that you meet the learning objectives while working with AI. / You may use generative AI freely on out-of-class work in this course, because your grade rests on what you can do in class without it. Use AI in the way that best prepares you to demonstrate your learning in person.]

What is permitted. You may use generative AI on any out-of-class assignment or work unless the prompt says otherwise. You are responsible for everything you submit. An AI tool’s error in your work is your error.

Verifying your understanding. In-class assessments [writing, seated exams, oral checks] are done without AI. Outside of class, assignments are designed so that your achievement of the learning objectives can be verified directly even when you have used AI. For all out-of-class work, I may ask you in writing or in a conversation to explain or extend your work, or otherwise verify your understanding of what you submitted.

Assignment-specific clauses

Adapt the following assignment-specific policies to fit your assignments. Each has a permission statement and an incentive for compliance. The incentive is what makes the clause enforceable, and it is specific to your course; the bracketed options are common choices. See our resource on Making Assignments AI-Resilient for more examples.

Reading and reading responses. You may use generative AI to explain a concept or a cited work as you read. [Choose one: Do not use it to summarize the reading/Do not use it to summarize the reading in place of engaging with the full text.]

Accountability: [Discussion of the reading in class is graded. / The reading will be the basis of a short in-class writing exercise or quiz. / Your reading response may be the starting point for an in-class exchange with me or a teaching fellow.]

Problem sets. Do not use generative AI to solve or check problems [including through code completion or a solver built into [tool]]. You may use it to explain a concept from the course or to pose practice problems. [Variant: You may use generative AI to check your work after you have attempted every problem. Note in your submission which problems you checked and what changed.]

Accountability: [A short in-class problem building on each problem set follows the due date. / Where the understanding shown in the in-class follow-up falls well short of the submitted work, the problem set grade will be marked down. / Problem sets are graded for completion, and the material is assessed on in-class quizzes and exams.]

Essays and research papers. You may use generative AI to [check grammar, find sources that you then read]. Do not use it to draft or revise your prose or to [devise a topic or thesis, create an outline, brainstorm counterarguments]. [Variant: You may use generative AI at any stage / Submit a process statement with your paper. / Cite your use of AI with a short statement of how you used it. /Include a link to your conversations with AI/Include a paragraph at the end of your essay describing how you used AI, what advantages AI provided, where AI fell short, and your overall experience using AI tools.]

Accountability: [Your topic proposal and outline are done in class. / After you submit, you will discuss your argument with me or a teaching fellow. / An in-class essay on a related question follows the due date.]

Programming assignments. Turn off code completion and AI chat in your editor for this assignment; instructions are on the course site. You may use generative AI to explain an error message or a language feature. [Variant: You may use AI coding tools. Submit your prompts or a short account of how you used them, and be prepared to explain and modify any part of your code in person.]

Accountability: [A short in-class coding exercise follows each assignment. / You will walk a teaching fellow through your code in section. / A portion of the assignment is completed in a supervised lab.]

Citation

Consider carefully whether and why you might ask students to cite their use of AI. If you require citation of AI, set a clear expectation for which uses require disclosure, and ask for the level of detail you will actually read.

Short statement. The student includes one or two sentences covering the required items. For example: "While outlining, I used Gemini to suggest search terms for finding sources on the labor history of the textile industry. I read the sources I found and cited them directly; nothing from the tool's output appears in the paper."

Full transcripts. Full transcripts, prompt logs, and document histories are also options. They are intrusive for students and slow to read, and they do not reveal unreported uses of AI. Ask for them when the student’s process is itself something you want to see and respond to.

Suspected violations

Start with a conversation about the work. Doing so gives the student a chance to respond to your concerns and gives you direct evidence of whether the submitted work reflects their understanding. Ask them to reconstruct a step, to explain a choice of sources or method, to define a term the submission uses and give a new example, or to describe what was hardest and what they tried first. A student who did the work can answer these questions readily.

A marked gap between submitted work and the student’s ability to explain that work in supervised settings provides evidence of reliance on outside help. Fabricated sources, quotations, or data, and artifacts an AI tool left behind are reliable evidence of AI-use. Neither your own sense that prose sounds like AI nor AI detection software is reliable evidence of AI use. See our guidance on Addressing AI-Related Academic Integrity Issues.