What colleges actually publish
The honest answer starts with what admissions offices put in writing, because that is what you are agreeing to when you press submit. None of the policies below mentions detection software. All of them mention consequences.
| Who | What the policy says | What it means for you |
|---|---|---|
| Common App | Submitting “the substantive content or output of an artificial intelligence platform, technology, or algorithm” as your own work is application fraud. | Binds every applicant to every Common App member college. Grammar and spelling review and early topic brainstorming are allowed. |
| Yale | Adopts the Common App policy; text composed by generating software “may result in admission revocation or expulsion.” | Says nothing about running detection software. The page is about the affirmation you sign. |
| Brown | AI “is not permitted under any circumstances in conjunction with application content,” except spelling and grammar review. | You affirm the application is your own work, factually true, and honestly presented. No detection tool is mentioned. |
The Common App fraud policy is the one that matters most, because it applies at more than a thousand colleges at once and the Common App runs its own fraud process on top of whatever each college decides. Yale’s AI policy page and Brown’s integrity page are the two clearest statements from individual schools, and both rest on the same mechanism: an affirmation that the work is your own, signed by you.
Are detectors actually being run on essays?
Quietly, in some offices, yes. Inside Higher Ed’s October 2023 report on a survey by Intelligent found that 50 percent of admissions offices were already using AI in their review process, that 60 percent of those used it to evaluate personal essays, and that some counselors were running essays through AI-powered detectors to spot machine-generated writing. No admissions official in that report, or on any major college’s admissions site, is on record saying their office runs a detector on every essay.
The scale of the problem they are reacting to is documented. A foundry10 survey reported by Education Week found that about a third of seniors applying for 2023–24 used an AI tool on their essays, roughly one in six for brainstorming or polishing, and about 6 percent had AI write the final draft. A 2026 Cornell and Carnegie Mellon study of tens of thousands of essays at one selective college found that essay language became significantly more uniform after AI tools launched, most of all among lower-income applicants, and that applicants who used AI more heavily were more likely to be rejected.
Two things are worth noticing in that last study. The researchers did not need a commercial detector to see the pattern; they measured homogenisation directly. And the applicants were not rejected by a tool. They were rejected by readers, for essays that sounded like everyone else’s.
Why the detector is not the real risk
It is tempting to treat this as a question of beating a scanner. It is not, for three reasons.
- Detectors get personal essays wrong in both directions. Short, polished, first-person prose is the hardest genre for any detector, and admissions readers know it. A flagged essay gets a closer human read, not an automatic rejection. Our guide to Turnitin false positives explains why edited, non-native, and formulaic writing trips detectors most.
- The human read is the detector. Your essay sits in a file next to your short answers, your recommendations, and sometimes an interview. A voice that appears in the essay and nowhere else is the signal readers actually use.
- You signed for it. The affirmation that the work is your own is what turns an AI-written essay into fraud. A detector score, high or low, does not change what you attested to.
Turnitin itself is largely absent from this picture. Its products page lists tools for coursework, assessment, and publishing, and nothing for admissions. Its AI detector is what you will meet once you are enrolled, on the first paper you submit through the LMS. What AI detector Turnitin uses covers how that one works.
What to do before you submit
- Keep your drafts. Version history in Google Docs or Word is the only evidence of authorship you will ever have, and the first thing anyone asks for.
- Use AI only where the policies allow. Spelling, grammar, and early topic brainstorming are fine under Yale and Brown’s wording. Sentences the tool wrote are not, even lightly edited.
- Read it against your short answers. If the essay uses vocabulary and sentence shapes that your 200-word responses do not, a reader will notice before any software does.
- If you did lean on AI, rewrite rather than rephrase. Paraphrasing tools produce exactly the uniform prose the Cornell study measured. Starting from your own notes is slower and it is the only fix that holds up.
- Then see how it reads. Run the finished essay through a detector once, to catch the case where your own writing happens to sound generated. Our free AI-writing estimate highlights the passages most likely to read as AI, and a real Turnitin AI Writing Report shows the score and highlights an institution’s detector would produce. The personal essay’s 650 words clear the 400-word minimum; short supplements do not.
If the report flags passages you genuinely wrote, that is useful information: it tells you which sentences a tired reader might doubt, and those are the ones to make more specific and more yours. The Turnitin AI score explained walks through how to read the percentage and the highlights.
Frequently asked questions
Does the Common App check essays for AI?
The Common App does not say it runs a detector. What it does is define fraud to include submitting the substantive content or output of an AI platform as your own work, and every member college can act on that definition. A report of fraud is handled through the Common App’s own process, which can include suspending your account, on top of whatever the college that noticed decides to do.
Does Turnitin check college application essays?
Turnitin’s products are sold to institutions for coursework, and its current products page lists nothing for admissions. An admissions office that wants a Turnitin AI score on an essay would need its own licence and a way to load application essays into it, and no major college publishes that it does so. If your essay is ever checked, it is far more likely to be a general-purpose detector or a human reader than Turnitin.
Can colleges tell if you used ChatGPT?
Sometimes, and not mainly with software. Admissions readers compare your essay with your short answers, recommendations, and interview notes; a 650-word essay in a voice that appears nowhere else in the file is the usual giveaway. A 2026 Cornell and Carnegie Mellon study of tens of thousands of essays at one selective college found that language became measurably more uniform after AI tools launched, and that heavier AI use was associated with rejection.
Is using Grammarly or spell-check on my college essay allowed?
Yes, under every policy quoted on this page. Yale states that using an AI platform to review grammar or spelling, or to seek general advice or topic suggestions at the start, does not constitute fraud, and Brown carves out spelling and grammar review explicitly. The line is drawn at substantive content: sentences and ideas that the tool wrote rather than corrected.
What happens if a college thinks my essay is AI-written?
It depends on the school, but the published consequences are severe: Yale says revocation of admission or expulsion, Brown lists revocation, expulsion, and revocation of credit or degree. In practice a suspicious essay usually gets a closer human read against the rest of your application rather than an automatic rejection. What you cannot do is argue the point later, because you affirmed the work was your own when you submitted.
Should I run my college essay through an AI detector before submitting?
If you wrote it yourself and want to know whether it reads as AI, yes, because false positives happen and a human reader can have the same reaction a detector does. If you used AI to write substantive parts, a detector is the wrong fix; rewriting in your own words is the only thing that satisfies the policies you are signing. Treat the score as a reading of how the essay comes across, not as permission.
Do AI detectors work on a 650-word essay?
Yes, 650 words is well above the minimum most detectors need. Turnitin’s AI detector requires at least 300 words, and our service needs 400, so the Common App personal essay qualifies while a 150-word supplement does not. Accuracy on short, polished, personal prose is lower than on long academic writing, which is exactly why a flagged score deserves a rewrite rather than a panic.