This page is about the essays you send with an application: the Common App personal essay, supplements, the UC personal insight questions and graduate statements of purpose. Essays written for a class once you are enrolled are a different question, covered on our page for students. Every policy below was read on the institution’s own site on 26 September 2026.
What the Common App says about AI
The rule sits in the Common App’s fraud policy, which defines fraud as “an intent to deceive, mislead, or otherwise perpetuate misinformation upon a Common App member regarding the candidacy of a student-applicant for admission.” Among the listed examples is this one, quoted in full:
“Submitting plagiarized essays or other written or oral material, or intentionally misrepresenting as one’s own original work: (1) another person’s thoughts, language, ideas, expressions, or experiences or (2) the substantive content or output of an artificial intelligence platform, technology, or algorithm;”
Two words carry the weight: “substantive” and “intentionally”. The clause targets presenting AI content as your own, not the existence of an AI tool somewhere in your process. What counts as substantive is left undefined, which is why the colleges below have written their own lists.
Every applicant also signs the application affirmation, certifying that everything submitted “is my own work, factually true, and honestly presented,” and accepting possible “admission revocation, expulsion, or revocation of course credit, grades, and degree” if that proves false. It does not mention AI by name. It does not need to.
On scanning, we found no Common App statement that it runs AI detection (fraud policy read 26 September 2026). The policy does say the Common App “monitors instances of alleged fraud” across its services, including “alleged fraud that Common App identifies”, without saying how it identifies anything. The only use of the word “detection” is about member colleges: they are not required to report every case, but “may do so consistent with the member’s own internal fraud detection protocols.”
The next paragraph limits what the Common App does with a college’s report. It “will not independently judge, evaluate, conduct any verification of, or perform an investigation on any information provided on or about a student-applicant and/or application from a Reporting Member,” except to address misuse of its own platform. That clause is about information a college sends in, not about applications in general. We did not read its Terms of Use.
When the Common App does find fraud, the policy lets it suspend or terminate the account and disclose the finding to the colleges on your “My Colleges” list. You then have ten business days from its written notice to appeal.
What individual colleges have published
Five colleges publish an AI rule for application essays; each was read on its own admissions pages. The policies range from Georgia Tech, which tells applicants who use AI to brainstorm, edit and refine with it, to Brown, which allows nothing past a spelling and grammar check.
| College | What its admissions office says | What it allows | Source |
|---|---|---|---|
| Caltech | All Fall 2027 applicants must review its AI guidelines before submitting supplemental essays, and “failure to comply … may result in the rescission of your admission to Caltech.” Listed as unethical: pasting from an AI generator, relying on AI to outline or draft, replacing your voice with AI content, and translating an essay written in another language. | Grammar and spelling review of a finished essay with tools such as Grammarly or Microsoft Editor; generating questions to start brainstorming; researching the application process. | Caltech: Undergraduate and Graduate Admissions AI Policy, guidelines for Fall applicants |
| Georgia Tech | A “Statement on AI” on its first-year review page: “you should not copy and paste content you did not create directly into your application,” and your “ultimate submission should be your own.” | If you choose to use AI, treat it as you would a human collaborator: “Use it to brainstorm, edit, and refine your ideas.” | Georgia Tech Undergraduate Admission: Application review process (Statement on AI) |
| University of California | “Content and final written text must be their own.” UC “conducts regular screenings to verify the integrity of the responses” and “may request authentication of the content or writing as the student’s.” | Advice on content and editing, including generative AI “to assist with readability”. | University of California: Statement of Application Integrity (PDF, last updated August 2023) |
| Brown | Adopts the Common App’s fraud definition and adds that AI use “is not permitted under any circumstances in conjunction with application content.” Essays and short answers “must be the work of that individual.” | Spelling and grammar review only, like any other proofreading tool. | Brown University Undergraduate Admission: Integrity in the application process |
| Yale | Submitting “the substantive content or output of an artificial intelligence platform” is application fraud, and statements “composed by text-generating software may result in admission revocation or expulsion.” | Grammar or spelling review, and “general advice or topic suggestions at the start of the writing process”. | Yale Undergraduate Admissions: AI policy |
The gaps between these rules are practical, not academic. Take one ordinary move: pasting a paragraph into a chatbot and asking it to tighten the wording. On our reading that fits Georgia Tech’s “edit, and refine”, and arguably UC’s allowance for readability help. It sits outside Brown’s spelling-and-grammar exception and close to what Caltech calls replacing your voice. One essay, recycled across a college list, can be within the rules at one address and outside them at the next.
Georgia Tech set out this position as early as July 2023, when Rick Clark wrote on the Georgia Tech admission blog: “you should not copy and paste directly out of any AI platform or submit work that you did not originally create.”
Among the selective colleges Kaplan polled, most had no policy at all. Kaplan’s survey reached admissions officers at 220 top national, regional and liberal arts colleges and universities, as compiled by U.S. News & World Report, by email between July and August 2025. On applicants using generative AI to write admissions essays, 68% said their college has no policy at all, 30% that its policy bans it and 2% that it allows it. That describes a group of selective colleges, not colleges in general. Kaplan’s 2024 survey, printed alongside for comparison, put the same figures at 72%, 25% and 2%. Where a college is silent, the Common App affirmation you signed still applies.
Two colleges that use AI to read essays
A different kind of statement is easy to mistake for detection. Two public universities say they use AI in reading applications, and neither of them describes it as checking who wrote the essay:
- UNC Chapel Hill says it “uses AI programs to provide data points about students’ common application essay and their school transcripts,” including writing style and grammar, so readers can focus on content (UNC admissions FAQ, read 26 September 2026).
- Virginia Tech announced on 24 July 2025 that each essay is now scored by one human reviewer and one AI reviewer, with a second human brought in if the two scores differ by more than two points (Virginia Tech News, read 26 September 2026).
Each describes AI helping to evaluate essays. None says it is looking for AI in them.
Do admissions offices run AI detectors?
We could not establish that any named college does, and we could not establish that none does. Here is what the sources support, and where they stop.
- None of the seven colleges above, in anything we read from them, says it runs AI-detection software on application essays.
- The University of California comes closest. Its Statement of Application Integrity promises “regular screenings” and says UC may ask you to authenticate your writing, but it does not say what the screenings use.
- Cornell’s and UBC’s graduate schools address detectors directly, and neither endorses them. Cornell says it does not rely on automated AI-detection software “as primary evidence” and judges concerns by review that is “not algorithmic screening alone”, wording that leaves open a detector as a secondary input. UBC’s Graduate School, in its general guidance on student AI use, quotes its Academic Integrity Hub, which strongly discourages detection software over concerns that include effectiveness, accuracy and bias.
- Kaplan’s survey asked about policies, not software. We found no survey by a named organisation, with a published sample and method, that measures how many admissions offices use detectors. Any figure you see for that is not one we could trace.
The honest answer for an applicant is that a given college might, you will rarely be told, and the rule you agreed to is the same either way.
Why a detector is a weak tool on a 650-word essay
The Common App personal essay runs 250 to 650 words, according to UNC’s admissions page, which puts it this way in its own guidance: “Remember: 650 words is your limit, not your goal.” We could not read a commonapp.org page that states the limit itself. The prompts are unchanged for 2026–27, and the Common App reports that the most popular choice in 2025–26 was “Topic of your choice”, at 28%.
Detectors publish minimums because they need volume to say anything. Turnitin’s guide to its AI Writing Report requires “at least 300 words of prose text in a long-form writing format” and says it does not detect short-form writing such as bullet points or tables. QuillBot’s detector needs 80 words and says “the longer the text the better the results.” We cite them because they publish their floors, not because any college told us it uses them. Why length matters so much is set out in our answer on text length and AI detection.
A full-length personal essay clears Turnitin’s floor, but not by the margin a long-form assignment would. Several pieces of an application do not clear it at all:
| Piece of the application | Word limit | Against Turnitin’s 300-word floor | Source |
|---|---|---|---|
| Common App personal essay | 250–650 words | Clears it, except at the bottom of the range | UNC Chapel Hill Undergraduate Admissions: Application prompts for 2026–2027 |
| Common App “Challenges and circumstances” | 250 words max | Below it | Common App blog: Announcing the 2025–2026 Common App essay prompts (27 February 2025) |
| Common App “Additional information” | 300 words max (cut from 650) | At the floor, at most | Common App blog: Announcing the 2025–2026 Common App essay prompts (27 February 2025) |
| UC personal insight question (answer 4 of 8) | 350 words max each | Just above it, if you use the full allowance | University of California: Personal insight questions |
| Caltech first-year supplements | Every prompt with a stated limit is capped at 200 words or less | Below it, for every capped prompt | Caltech: Supplemental application essays |
The genre, which is our reasoning rather than a measured finding. Beyond length, the personal essay is a hard case for a detector. It is written in plain first person, tends to follow a familiar arc from a moment to what it taught you, and is often read and tidied by a counsellor, a teacher and a parent before it is sent. Each of those pushes prose toward the even, conventional surface that detectors associate with machine text. Applicants writing in a second language face the added risk the research documents, which is the reason non-native English writers are flagged more often and the reason Cornell gives for not leaning on detectors. We have not tested detectors on admissions essays, so treat this paragraph as an argument, not evidence.
Supplemental essays and short answers
Supplements are where the length problem bites. Caltech’s 2026–27 first-year prompts that state a limit run from 100–150 words to a maximum of 200, which puts every capped prompt below Turnitin’s floor. The UC personal insight questions ask for four answers of up to 350 words each: four separate short texts, each only just over 300 words at full length. The Common App’s own short fields are short too. Its 2025–26 announcement cut “Additional information” from 650 words to 300, and kept “Challenges and circumstances” at 250.
A short answer gives a detector almost nothing to measure, so a score on one tells you very little in either direction. The rules are not relaxed for short texts, though. Brown’s ban covers “short-answer questions” by name, and Caltech wrote its guidelines specifically for its supplemental essays. Short means hard to measure. It does not mean outside the policy.
Graduate applications and statements of purpose
A statement of purpose is judged by the university that receives it, wherever the applicant lives. An applicant in India, Nigeria or the Philippines applying to a US, UK or Canadian programme is bound by that programme’s rule, not by any rule at home. Four graduate policies, each read on the institution’s own site:
| Institution | What it requires | On detection | Source |
|---|---|---|---|
| Cornell Graduate School (US) | Any GenAI use, even a permitted one, must be disclosed in a short “AI Use Disclosure” added to the statement of purpose or personal statement. Using AI to translate your writing into English “in whole or in substantial part” is on the prohibited list. | Says it does not rely on automated AI-detection software as primary evidence, citing false positives for non-native English speakers. | Cornell University Graduate School: Use of generative AI in graduate admissions applications (updated 1 September 2026) |
| University of Oxford (UK) | AI may help you research, “but the ideas and their expression must be yours.” A personal statement or research proposal must be entirely your own work. “Where you use AI to assist in initial research, you should declare how you have used it in a footnote.” If misuse is found after an offer, “the offer may be withdrawn.” | Names no detector. Says academic assessors judge whether AI was used inappropriately, and that interviewers may probe your submitted work in detail. | University of Oxford: Using artificial intelligence in your graduate application |
| UBC Graduate School (Canada) | Applicants certify “that the ideas represented in responses are my own,” and any GenAI-produced text in the application “must be cited, and the process used made transparent.” | Nothing in its applicant section. In its general guidance on student AI use, it quotes UBC’s Academic Integrity Hub: detection software is “strongly discouraged at this time”. The applicant section says programmes may ask applicants to express their ideas in writing and orally during admission. | UBC Graduate School: Use of generative AI (graduate applicants) |
| Caltech (US, graduate and undergraduate) | The same policy as its undergraduate row above, approved by both its Undergraduate Faculty Admissions and Graduate Studies committees for the Fall 2027 cycle. | Does not mention detection. | Caltech: Undergraduate and Graduate Admissions AI Policy, guidelines for Fall applicants |
Cornell’s wording is worth reading in full if you are applying anywhere, because it is the clearest published explanation of why an admissions office might distrust a detector: such tools “can produce inequitable false positives, particularly for non-native English speakers and other writers whose style differs from the norms built into detection software.” It says concerns are evaluated through “holistic, human-centered review, not algorithmic screening alone.” Oxford takes a different route to the same place: it names no tool, and says interviewers may probe your submitted work in detail.
Translation is the rule international applicants most often miss. Caltech and Cornell both name it. Caltech lists “translating an essay written in another language” as unethical use, and Cornell prohibits using AI to translate your writing into English in whole or in substantial part. If you draft in your first language, the English version has to be yours too, at those schools at least. HumanFlow works in English only, and nothing it does changes what these policies allow.
Disclosure is the other thing graduate rules add. Cornell, Oxford and UBC each require it in some form: Cornell an AI Use Disclosure in the statement for any use at all, Oxford a footnote declaring how AI helped with initial research, and UBC a citation for any GenAI-produced text. A sentence of disclosure costs little. Leaving one out where it is required turns a permitted use into a policy breach.
Using AI without misrepresenting yourself
The rules above agree on more than they differ. Every one of them allows something, and every one draws the line at submitting text a machine wrote as your own. A workable approach:
- Read each college’s own rule for every school on your list, and follow the strictest one if you are sending the same essay to all of them.
- Keep your drafts, notes and version history. UC says it may ask you to authenticate your writing, and a dated trail of drafts answers that better than any score. How to show you wrote it covers what to keep.
- Disclose where you are asked to. For graduate programmes such as Cornell’s, a missing disclosure is itself the breach.
The bigger risk in an application is not a detector at all. It is invented detail, which chatbots supply readily and which becomes a misrepresentation once you sign the affirmation; our personal statements guide explains why that danger is sharper here than anywhere else. Where the line sits between help and misconduct more generally is the subject of whether using an AI humanizer is cheating.
Where HumanFlow fits
We make an AI detector and a rewriting tool, so we have an interest in this subject and you should weigh what we say with that in mind. Our detector works in English only, and the free plan scans up to 1,500 words at a time. It returns our model’s estimate, which says nothing about what any college would conclude, since none of them names a tool.
On the rewriter, the policies above answer the question for us. Brown allows nothing past spelling and grammar; Caltech bars replacing your voice with AI-generated content; Yale treats submitting AI output as fraud. Rewording your application essay with an AI tool, ours included, is not among the permitted uses at any of those three, and we would not recommend it for an application at all.
Applying somewhere else?
Applying to UK universities through UCAS is a different system with its own screening and declaration, covered in our guide to UCAS personal statements and AI. Taking AP Seminar or AP Research? College Board runs its own checks on that work, and our answer to “Does College Board check for AI?” sets them out. For essays written for a class once you are at college, see our guide to essays.
Limits of this page
Every row is an institution describing its own rules, read on one date. Admissions policies are rewritten between cycles, and Caltech’s is explicitly for Fall 2027. We read statements from ten institutions out of thousands (seven colleges and three further graduate schools, with Caltech counted once because one policy covers both), chose them because they publish a statement, and could not capture Duke’s or Princeton’s on the day, so this is not a sample of what colleges typically do. A college that publishes nothing about detection may still use it.