Plagiarism and AI at Stanford: AI is treated like another person

The short answer

Stanford did not write a new AI rule. It decided that using generative AI counts as getting help from another person — and let its existing Honor Code do the rest.

That is the most useful test in this entire series, because you can apply it yourself, to any task, without reading a policy.

The analogy rule

From the Board on Conduct Affairs guidance, adopted 16 February 2023:

“Absent a clear statement from a course instructor, use of or consultation with generative AI shall be treated analogously to assistance from another person.”

Everything follows from that. The Honor Code already prohibits “giving nor accepting unpermitted academic aid”. If a person could have helped you that way, the AI may. If a person could not, it may not.

Apply it to the cases students actually worry about:

• Explaining a concept you found hard. A friend could do that. So could a tutor. Fine.
• Reading your draft and pointing out a weak paragraph. Depends on the course — the same as asking a classmate would.
• Writing a paragraph for you. A person doing that would be unpermitted aid in almost any assessment. So is the AI.
• Answering the exam question. Obviously not — and Stanford says so directly: “Students should not use AI to substantially complete an assignment or exam (e.g. by entering exam or assignment questions).”

Instructors keep control: “Individual course instructors are free to set their own policies regulating the use of generative AI tools in their courses, including allowing or disallowing some or all uses of such tools.” The analogy is the default, not the ceiling.

Note also what Stanford did notdo. Its Honor Code violations list — copying, unpermitted collaboration, plagiarism, giving or receiving unpermitted aid, “representing as one’s own work the work of another” — does not mention AI at all. It did not need to.

Default to disclosing

The disclosure rule is short, and its tie-breaker is the part worth remembering:

“Students should acknowledge the use of generative AI (other than incidental use) and default to disclosing such assistance when in doubt.”

Two things are unusually sensible here. There is an incidental use carve-out, so nobody expects you to declare autocomplete or a spell-check. And where you genuinely cannot tell which side of the line you are on, Stanford tells you which way to err — disclose.

That is good advice well beyond Stanford. Declaring a permitted use costs you nothing; failing to declare one that needed declaring turns an allowed practice into a violation.

Stanford publishes no required citation format, so make the acknowledgement clear and specific: which tool, for what, on which part. Our AI disclosure statement guide covers the shape.

Detection needs advance notice

Stanford’s position here is different from most in this series, and worth stating precisely. It does not ban detectors and it does not disparage them. It attaches a condition:

“OCS recommends course instructors provide clear advance notice that they may use detection software to review work submitted for use of generative AI.”

That is a procedural fairness rule rather than a technical judgement: if your work is going to be run through something, you should have been told beforehand. It is a reasonable standard, and if you are ever facing a case built on detector output it is fair to ask whether that notice was given.

Whether Stanford centrally licenses any named detector is not publicly documented, and no similarity vendor is named on the Honor Code or guidance pages. So we cannot tell you what, if anything, runs.

Whatever tool is used, the limit is the same and we apply it to ourselves: our published error rates are measured on English and German corpora, and no detector — ours included — produces proof. See how accurate AI detectors really are and, if a case has begun, what to do when you are accused.

If a detector score is what started this, our guide to an AI detector flagging your thesis sets out what that number does and does not establish on its own.

What the Honor Code asks of instructors

Stanford’s Honor Code is unusual in being reciprocal. It places duties on staff as well as students, and those duties are worth knowing because they are yours to rely on:

“Instructors will support this culture of academic honesty by providing clear guidance, both in their course syllabi and in response to student questions, on what constitutes permitted and unpermitted aid. Instructors will also not take unusual or unreasonable precautions to prevent academic dishonesty.”

So if your syllabus is silent on AI, asking is not an imposition — the Honor Code says answering is part of the instructor’s job. Ask, and keep the reply.

The current text applies to incidents on or after 1 September 2023; Stanford keeps the previous version published for earlier cases. If something you did falls before that date, check which text applies.

Proctoring changes in 2026

This is the freshest governance change in this series and it changes the exam experience directly:

“Through the end of Academic Year 2025–26, proctoring is not required and is only permitted in courses participating in the Academic Integrity Working Group (AIWG) proctoring pilot. Beginning September 1, 2026, proctoring of in-person assessments will be permitted university-wide.”

The route it took is on the record: the Board on Conduct Affairs, the Undergraduate Senate, the Graduate Student Council, the Senate of the Academic Council and the Office of the President all voted for it, and on 22 May 2026 the BCA “adopted the Stanford Proctoring Bylaw”, which “provides a framework for fair and effective in-person proctoring”.

It is worth reading that alongside the Honor Code’s own instruction that instructors “will not take unusual or unreasonable precautions” — the Bylaw is the framework meant to keep proctoring inside that limit.

Two practical points. Expect proctored in-person assessment where you previously did not. And note the direction of travel: Stanford’s answer to AI is not more detection but more supervised assessment.

What Stanford does not publish

We would rather list the gaps than fill them, and for a university this heavily written about the temptation to fill them is exactly the problem. On the reachable pages we could not establish:

• The penalty scale. The Honor Code page defers sanctions to the Student Conduct Penalty Code Bylaw, which we did not retrieve. We are not going to describe a ladder we have not read.
• The process. The charging path, hearing panel composition, appeal route and standard of proof are not documented on the pages we could reach. The Office of Community Standards administers cases and the Board on Conduct Affairs sets policy — that much is clear.
• Any similarity vendor, self-check route, or retention statement.
• Any case counts.Stanford’s statistics address returned a 404, so no figures are given here. If you see a number attributed to Stanford, ask where it came from.

If a case concerns you, the Office of Community Standards is the place to ask for the Penalty Code Bylaw and the applicable procedure in writing.

Before you submit

• Apply the person test.Would a human doing this for you have been permitted aid? That is Stanford’s rule in one question.
• Ask your instructor if the syllabus is silent. The Honor Code makes answering part of their role.
• Disclose when in doubt. Stanford tells you which way to err.
• Never put an exam or assignment question into a tool. That is named explicitly.
• Keep drafts and version history. With no published detector position, a question about your work is answered by your work.

If you want a read on your own text before you hand it in, our AI check is free up to 1,500 words, with no account and no name.

Sources

• Stanford University: The Honor Code — current text applies to incidents on or after 1 September 2023; carries the dated 2026 proctoring updates.
• Stanford University: Generative AI policy guidance — adopted 16 February 2023 by the Board on Conduct Affairs.

This page is orientation, not legal advice. What binds you is your course syllabus and the Stanford Honor Code applicable to your incident date.