Almost every guide to this question tells you what colleges think about AI in general. Very few quote what specific schools actually published, and some of the summaries in circulation are simply wrong.
So here are the policies themselves, quoted and linked. The short version is that schools agree less than you would expect — and they disagree most sharply about the thing students most want to know.
The floor: what applies everywhere
Whether or not a school has published its own policy, the Common Application's fraud policy applies. It defines application fraud to include misrepresenting as your own original work the substantive content or output of an artificial intelligence platform.
Read carefully, that prohibits submitting generated text. It says nothing about brainstorming, thinking out loud, or getting feedback on something you wrote. For the schools below that have published nothing of their own, this is the entire rule.
The single most useful thing to know: Cornell lists brainstorming essay topics as an ethical use of AI. WashU lists it as an unethical one. Their policies are otherwise nearly identical in wording.
At a glance
| School | Building a school list | Brainstorming topics | Drafting or writing | Grammar & spelling |
|---|---|---|---|---|
| Brown | — | No | No | Yes |
| Caltech | Yes | Yes | No | Yes |
| Cornell | Yes | Yes | No | Yes |
| Emory | — | — | No | Yes |
| Harvard | — | — | No | — |
| Johns Hopkins | — | Yes | No | Yes |
| Northwestern | Yes | Yes | No | Yes |
| Penn | — | Yes | No | Yes |
| Swarthmore | — | Yes | No | Yes |
| WashU | Yes | No | No | Yes |
| Yale | — | Yes | No | Yes |
| Amherst | — | — | — | — |
| Columbia | — | — | — | — |
| Dartmouth | — | — | — | — |
| Duke | — | — | — | — |
| MIT | — | — | — | — |
| Princeton | — | — | — | — |
| Stanford | — | — | — | — |
| Williams | — | — | — | — |
A dash means the school's published policy does not address that activity — not that it is permitted or prohibited. Where a policy is silent, the Common Application fraud policy is what applies. The four columns also flatten some real distinctions: Cornell prohibits using AI to outline an essay while permitting brainstorming, Northwestern prohibits submitting AI's writing produced from an outline you supply, and Emory permits organizing ideas within a personal statement.
The policies
Cornell University
Ethical uses of generative AI include researching colleges, brainstorming essay topics, and reviewing the grammar and spelling of your completed essays.
Cornell then lists outlining, drafting, and writing essays as unethical, along with translating an essay written in another language. The outlining prohibition is unusual — no other school in this set draws that line.
Washington University in St. Louis
Unethical uses include (but are not limited to) using AI to brainstorm essay topics or draft or write your essays.
WashU organizes its guidance into always, sometimes, and never. Researching colleges and building a college list are in the always column. Grammar and spelling review of completed essays is in sometimes. Brainstorming essay topics is in never — the only school here to place it there.
Brown University
The use of artificial intelligence by an applicant is not permitted under any circumstances in conjunction with application content.
Brown then carves out one exception: AI may be used for spelling and grammar review, in the same way as any other proofreading platform. The policy also states that the content of all essays and short answers must be the work of the individual. Read strictly, this is the most restrictive policy in this set.
Emory University
AI tools could polish or refine, but they should not replace or generate your personal input or creativity.
Emory's guidance is principles-based rather than a list of permitted activities. It is headed "AI as a Writing Coach, Not a Creator," and it explicitly permits using AI to clarify your thoughts in the Activities section or organize ideas in a personal statement. It also notes that family, teachers, and counselors remain the best coaches.
Yale University
Using an AI platform to review one's grammar or spelling, or to seek general advice or topic suggestions at the start of the writing process does not constitute application fraud.
Yale draws the line at generated text. Submitting the substantive content or output of an AI platform is application fraud under Common Application policy, and Yale states plainly that submitting personal statements or other written responses composed by text-generating software may result in admission revocation or expulsion.
Yale is also the only school here to decline to recommend AI either way, noting that some applicants may find these tools useful and others may not. Its policy permits without endorsing.
Caltech
Ask yourself whether it would be ethical to have a trusted adult perform the same task you are asking of ChatGPT.
Caltech works that test through: a teacher reviewing your essay for grammar and spelling is obviously fine; a teacher writing a draft for you to tweak and submit is obviously not. Its ethical list names generating questions or exercises to kick start brainstorming, grammar and spelling review of completed essays, and researching the application process. Its unethical list names copying from an AI generator, relying on AI content to outline or draft, replacing your voice with AI content, and translating an essay written in another language.
Caltech is the only school here to attach an explicit consequence to its AI guidelines specifically: failure to comply may result in rescission of admission. It also states the guidelines will change as the technology does, and applies them to graduate as well as undergraduate applicants.
Johns Hopkins University
You may use AI tools for brainstorming or refining ideas before the writing process begins and for basic proofreading of your own writing.
Hopkins adds that all material submitted within the application must be your own original work, and that submitting the substantive content or output of an AI platform is application fraud under Common Application policy.
Hopkins is the only school here to reproduce the Common Application affirmation in full, including what a breach costs. Applicants certify that everything submitted is their own work, factually true, and honestly presented — and acknowledge they may face admission revocation, expulsion, or revocation of course credit, grades, and degree if it isn't.
Northwestern University
Constructive uses of generative AI… brainstorming essay topics.
Northwestern's constructive list also covers researching colleges, building a tour itinerary, exploring how academic interests connect to fields of study, and reviewing grammar and spelling of completed essays. Its problematic list covers submitting AI output as your own writing, translating an essay from another language, handing AI an outline and submitting the finished piece, and reworking a supplemental essay written for a different school.
Northwestern is the only school in this set to name equity as a reason for permitting AI at all, noting the guidance is particularly relevant for students in settings where college-prep resources and support may be limited. It is also narrower than Cornell on outlining: Cornell prohibits using AI to outline, while Northwestern prohibits submitting AI's writing built from an outline you supplied.
University of Pennsylvania
Getting help is not the same as losing ownership.
Penn permits talking an idea through with a parent, teacher, or counselor, having someone proofread, and using an AI tool to organize your thoughts or check grammar — none of which, in their words, compromises your application certification. What would compromise it is submitting work, ideas, or experiences that aren't authentically yours.
Penn offers the clearest test any school has published: if you couldn't explain in a conversation how you arrived at what you wrote, or speak to it in your own words, it isn't ready to submit. Penn is also the only school here to disclose its own AI use — administrative tasks like completeness checks and transcript scale interpretation, with a staff member confirming every outcome, and no evaluative decisions made by AI.
Harvard University
We are aware that it is possible to use tools or other services to fraudulently create essays or pass off another's work as one's own.
Harvard states that doing so violates both the Common Application standards and the Harvard College Honor Code. The guidance is aimed at authorship of submitted material and does not address brainstorming, feedback, or proofreading either way — so on those activities, the Common Application policy is what applies.
Swarthmore College
Any assistance with writing comes at the beginning or end of the process — such as brainstorming ideas for essay topics or asking for feedback on readability, spelling, and mechanics.
Swarthmore frames the question in terms of when help arrives rather than who provides it, applying the same standard to AI and to another person. Help at the start and at the end; the writing itself in between is yours.
Princeton University
Princeton has not published an admissions AI policy. Its Dean of Admission has addressed the question publicly on the admissions site, which is informative but not a stated rule.
Stanford University
Stanford has no unified admissions AI policy. Individual graduate programs publish their own — the Knight-Hennessy Scholars policy is one example — but these do not govern undergraduate applications. The Common Application fraud policy is what applies.
Amherst, Columbia, Dartmouth, Duke, MIT, and Williams
None of these has published an institution-wide AI policy for undergraduate admissions. MIT writing faculty have commented publicly on AI-assisted writing, but faculty commentary is not an admissions rule. For all six, the Common Application fraud policy is the operative standard and the only one.
That six schools of this caliber have published nothing is itself worth noticing. Silence is common, and it is not the same as either permission or prohibition.
What to do with this
Three things worth taking away.
- Check the schools on your own list. There is real variation across schools, and a similar-sounding policy that permits brainstorming at one institution may prohibit it at another. Your list is the only one that matters.
- Every school prohibits the same thing. Submitting text you did not write. That is the line nobody disagrees about, and it is the one with consequences — the Common Application affirmation you sign carries admission revocation and expulsion as stated penalties.
- Silence is not permission, nor is it prohibition. Where a school has published nothing, the Common Application policy applies and nothing more.
We build a tool that helps students find what to write about, and we have an obvious stake in how you read all this. So we will say the uncomfortable part out loud: if a school on your list says not to use a tool like ours for brainstorming, don't. That includes ours. We wrote separately about how to evaluate any AI essay tool, including the ones that advertise their ability to evade detection.
How this was compiled and what it is not
Sources. Every policy above was read directly from the school's own website, and every entry links to that page. Quotations are verbatim. Where a school's language is principles-based rather than a list of activities, we have said so rather than forcing it into the table.
This is not legal or admissions advice. It is a reading of published documents, and reasonable people read some of them differently — Brown's in particular. Where a policy is ambiguous, we have quoted it rather than resolved it. If your reading differs from ours, the source is one click away, and the school's own words govern.
Policies change. Several of these were published or revised within the past year, and schools that have nothing today may publish tomorrow. The verification date is at the top of this page. If you find something out of date or wrong, let us know — hello@youward.co.