Is Using AI Cheating? How to Read Your Professor's AI Policy

Is using AI cheating? The honest answer is that it depends on the course, and sometimes on the assignment. Using AI counts as cheating when it breaks the rule your instructor set for that piece of work, or when you hand in what an AI produced as if you'd made it yourself. The same request to a chatbot can be encouraged in one class and count as misconduct in the next.

The catch is that the rule isn't always where you'd look, and some courses never state one.

Is using AI cheating? It depends on the class

Academic integrity rules have long banned unauthorized help and passing off someone else's work as your own. Generative AI didn't change that; it made the line move from class to class. At many universities, including Stanford and Harvard, your instructor draws it, one course at a time.

• Stanford has a default for courses that say nothing about AI. AI help is treated like help from another person, and using AI to substantially complete an assignment or exam, for example by entering the questions, isn't allowed. Students should disclose any use beyond the incidental, and disclose when in doubt. Instructors are free to set their own rules instead (Stanford Office of Community Standards).
• At Harvard, all faculty must tell students their course's AI policy, and the Office of Undergraduate Education asks them to post it on Canvas. Its sample policies remind students that classes may set different rules and that following each one is the student's job (Harvard Office of Undergraduate Education).

Most students learn the rules inside the course. In an Inside Higher Ed and Generation Lab survey, 41% of students said they knew when AI was allowed from syllabus statements and 35% from instructors covering it in class; only 17% credited a policy their college had published. In August 2025, 13% said they didn't know when they could use AI for coursework, down from 31% in spring 2024 (Inside Higher Ed).

In May 2026, Science published what UC Berkeley calls the largest study yet of undergraduate AI use: more than 95,000 students at 20 public research universities, surveyed in spring 2024. About two-thirds used generative AI, and at least 9% of those users said they'd used it to cheat: 26% of daily users, against 7% of monthly users. The study's lead author, UC Berkeley's Igor Chirikov, calls the numbers conservative, partly because students first had to recognize when a use wasn't allowed. That's hard when course policies, in his description, range from allowing AI everywhere, exams included, to banning it outright (Berkeley News).

The four kinds of course AI policy

Universities publish sample AI statements for instructors to adapt. They fall into four types:

How to read the AI statement in your syllabus

Start with the syllabus, usually under academic integrity or course policies. Then check the Canvas assignment page, the rubric and announcements: a rule for one assignment is more specific than the course default. Cornell's teaching center tells instructors to state AI rules in the syllabus, in assignment instructions and out loud in class, so the policy can live in any of those places (Cornell Center for Teaching Innovation).

Then answer five questions:

1. Which tools does it cover? "AI" can mean more than chatbots. The University of Maryland's sample statements list Grammarly next to ChatGPT and Gemini (University of Maryland). If your word processor, grammar checker or search engine has AI built in, ask whether it counts.
2. Which stages? Brainstorming, research, outlining, drafting, editing, coding and studying can each be treated differently.
3. Which assignments? Check whether one rule covers the whole course or each assignment sets its own.
4. What do you have to disclose, and how? A sentence at the end, a citation in APA or MLA, a link to the chat, or a full transcript in an appendix.
5. What happens if you break it? Harvard's samples all say a violation will be treated as academic misconduct.

A worked example

Take the Maryland sample from the table: no AI for brainstorming or research, AI allowed for writing and revising once your ideas are formed, and a note naming the tool and how you used it. In practice:

• Allowed: asking AI to point out unclear sentences in a paragraph you wrote, or to suggest tighter wording for your draft.
• Not allowed: asking AI for paper topics, for sources, or to summarize the research for you.
• Ask first: an outline. Is building one brainstorming or writing? The statement doesn't say, so get an answer in writing.
• Disclose: a note naming the tool and describing the edits it helped with.

Under a policy like this one, the research summary is yours to write. If that's a skill you haven't practiced, The Word on College Reading and Writing, a free course on Lykke, has a section on summarizing and critiquing that separates an objective summary from your own analysis.

What to do if your syllabus doesn't mention AI

1. Check for a school default. Search your school's academic integrity or honor code pages for "generative AI." Stanford's rule above is one example: with no course policy, AI counts as outside help.
2. Don't read silence as permission. If the syllabus is quiet and your school has no default, you don't know the rule yet. Michigan State's ombuds office reminds students that ignorance of the rules is not a defense (MSU Office of the University Ombudsperson).
3. Ask in writing, about one use on one assignment. "Can we use AI?" gets a vague answer. "Can I use AI to check grammar in my own draft of the Unit 2 essay?" gets a useful one.
4. Keep the reply. If the answer comes in class or office hours, send a two-line email confirming what you understood, so it's written down.
5. Ask again when the task changes. Permission to use AI for grammar on an essay doesn't cover a lab report.

A short email is enough:

Subject: AI use on [assignment name], [course number]

Hi Professor [last name],

I couldn't find an AI policy for [assignment name] in the syllabus or on Canvas. May I use an AI tool to [the specific use, such as checking grammar in my own draft]? If so, how would you like me to disclose it?

Thank you,

[your name]

AI uses that are usually fine, and ones that usually aren't

These are where lines tend to fall, not rules. If your syllabus says something different, the syllabus wins.

Usually fine for studying

None of these produces work you hand in, though a strict course policy can reach further, so check yours:

• Having AI quiz you on your own notes, the practice testing our midterm study plan is built on.
• Asking for a second explanation of a concept after the lecture or the reading.
• Generating extra practice problems, then checking the answers against your textbook or the posted solutions.
• Turning your notes into flashcards.

Chirikov, the Berkeley researcher, suggests a gut check whenever you use AI: could you explain this without the tool, could you do a similar task on your own tomorrow, and did AI help you understand the material or mainly help you finish faster?

Ask first

• Brainstorming topics, angles or counterarguments.
• Outlining.
• Grammar, tone and clarity edits, including grammar checkers and the AI built into word processors.
• Translating your own writing.
• Explaining or debugging code for a programming assignment.
• Feedback on a draft.

If the answer is no, you can learn to do most of these on your own. The free Online Writing Lab course has sections on prewriting and idea generation, from freewriting to mapping, and on drafting, revising and editing in separate passes. For code, the free Introduction to Computer Science and Programming in Python course has a section on testing and debugging.

Usually not allowed

• Pasting assignment or exam questions into a chatbot and turning in what comes back. Stanford's default rule names this exact case.
• Handing in AI-written text as your own words. APA's style team says submitting AI-generated work as though you did it yourself counts as academic misconduct for students (APA Style).
• Using AI during a quiz or exam unless the instructor says you can.
• Citing sources an AI suggested without reading them. MLA warns that AI tools can make up sources or misstate what a source says, and recommends citing the sources themselves (MLA Style Center).
• Any AI use in a course that bans it, including "just editing" when the ban covers editing tools.

When a source does belong in your paper, read it and put it in your own words. The free English Composition II course's section on quoting and paraphrasing shows how, including where a close paraphrase turns into patchwriting.

How to disclose and cite AI in APA or MLA

Disclose AI use the way your instructor asks. If they don't specify, a short note plus a citation in your course's style covers both. APA says that if you used AI while writing a paper, you'll usually need to disclose the use and cite the chat, the tool or both.

Write a disclosure note

Say which tool you used, what you used it for and what you did yourself. For example:

I used ChatGPT to suggest counterarguments to my thesis and to flag unclear sentences in my second draft. I chose which counterarguments to answer, and the final wording is mine.

APA (updated September 2025)

• A specific chat, when it helps your reader: Company. (Year, Month Day). Title of chat [Generative AI chat]. Tool or model name. URL of the shared chat. For a ChatGPT chat from this fall, the in-text citation is (OpenAI, 2026).
• The tool in general, for example when you used it to edit your own writing. The year is when the tool was last updated: OpenAI. (Year). ChatGPT [Large language model]. https://chatgpt.com/
• The company is the author; APA says AI can't be one.
• Describe your prompts in your text or in an appendix, not in the reference.
• Skip version numbers, and add the model name if you know it.

MLA (updated August 2025)

• Don't list the AI as the author. Describe the prompt where the title goes, then give the tool, the model, the company, the date and a link: “Explain the causes of the Dust Bowl” prompt. ChatGPT, model [model name], OpenAI, 28 Sept. 2026, followed by the chat's share link.
• Use a stable share link to the chat if the tool offers one, and the tool's general address if it doesn't.

Both styles now point to shared chat links, so save them as you go.

New to APA and MLA in general? English Composition II, a free course on Lykke, has a section on citation standards in MLA and APA that covers in-text citations and reference lists in both, and its Citation Transformer demo switches a book, a journal article and a website between the two styles.

Can professors tell if you used ChatGPT? What AI detectors show

Sometimes, but rarely from a detector score alone. A detector estimates how much of a text looks machine-written; it can't see how you worked. Professors lean more on the work itself, and on asking you about it.

• Turnitin shows only an asterisk, with no percentage, when it detects AI in 1% to 19% of a submission, to avoid false positives. It says submissions under 300 words are likely to get a less accurate score, and that an AI score should not be the sole basis for action against a student. It also says it flags AI text reworded by paraphrasing tools and, since August 2025, by "bypasser" tools built to disguise it (Turnitin).
• OpenAI launched its own AI text classifier in January 2023 and withdrew it on July 20, 2023, citing its low accuracy. In OpenAI's evaluation, it caught 26% of AI-written text and labeled human writing as AI-written 9% of the time (OpenAI).
• Detectors have misjudged non-native English writers. In a 2023 study by Stanford researchers, seven widely used detectors labeled, on average, 61% of 91 TOEFL essays by non-native English speakers as AI-generated, while classifying essays by US eighth graders almost perfectly (Liang et al., Patterns).
• Schools differ on detectors. Harvard's Faculty of Arts and Sciences advises instructors not to count on automated detection and doesn't plan to provide a detector. Cornell's teaching center doesn't recommend detectors for integrity cases because they can't provide definitive evidence. Stanford's conduct office recommends that instructors tell students in advance if they may use detection software.

Cornell also tells instructors to warn students that they may be asked to explain their work out loud, and lists the evidence to look for: citations to sources that don't exist, methods far beyond the course, a student who can't explain or defend their own submission, and work that doesn't match earlier drafts or assignments. At Cornell, that evidence has to meet a "clear and convincing" standard.

So your best protection is a visible process. Write in Google Docs, or in Word with the file saved to OneDrive, so version history exists. Keep your notes, outline and sources, and know your paper well enough to talk through any part of it.

What to do if you're accused of using AI

An accusation is stressful, but it starts a process; it isn't a finding.

1. Read the message carefully. Note what you're accused of, which policy it cites and any deadline. Don't delete drafts, files or chats. In Google Docs, deleted version history can't be recovered.
2. Find your school's process. Search your school's site for academic integrity or student conduct, and read the steps and deadlines. At Michigan State, for example, a student who disputes a charge meets the instructor first, can take it to the department chair next, and can then request a hearing before a university board (MSU Office of the University Ombudsperson).
3. Gather your process. In Google Docs, click Last edit at the top right to see earlier versions and who changed what (Google Docs Editors Help). Only people who can edit a Google Doc can see its version history, so be ready to show it from your own account. In Word, open version history from the file name or the File menu; it only works for files stored in OneDrive or SharePoint (Microsoft Support). Add your outline, notes, the sources you read, earlier drafts and any email approving your AI use.
4. Meet your instructor. Ask what raised the concern. Walk through how you wrote the work and offer to answer questions about it. If English isn't your first language and a detector flagged you, the Stanford study above is worth raising.
5. Get a guide to the process. An ombuds office is a confidential place to understand your options; Michigan State's advocates for fairness and due process rather than for either side. Your dean of students office and academic advisor can also explain the steps.
6. Be accurate about what you did. If you used a tool the policy allowed, such as a grammar checker in a course that permitted editing help, say so and show the policy text or the email.

A 60-second check before you use AI on graded work

• Have you read the course policy and this assignment's instructions?
• Does the policy cover this tool and this stage of the work?
• If anything is unclear, have you asked in writing and saved the answer?
• Are you keeping a trail: drafts, version history and chat links?
• Could you explain every part of what you're handing in without the tool?

If any answer is no, sort that out before you start.