The Common Application's fraud policy treats misrepresenting as your own the substantive content or output of an artificial intelligence platform, technology, or algorithm as application fraud. The stated consequences are admission revocation or expulsion. Not a warning, not a lower score — revocation. As the fall application season begins, that single sentence is the most expensive thing a student can misunderstand about AI this year, because the line it draws is narrower and stranger than almost anyone assumes. The operative word is misrepresenting: the fraud is passing AI output off as your own writing, not the mere fact of having used a tool.
Here is what makes this genuinely hard rather than obvious: the schools do not agree with each other on the details, they publish those details in different places, and at least one prohibits something most students would never guess was banned. This is a guide to what the actual published policies say, quoted from the schools themselves, and how to use AI in a way that keeps you safely inside all of them.
Quick summary: Common App treats AI-generated substantive content as fraud, with consequences up to admission revocation or expulsion. Yale permits using AI to review grammar and spelling and to seek general advice or topic suggestions at the start of the writing process, but says the substance must be yours. Caltech publishes an explicit permitted and prohibited list, and warns that failure to comply may result in rescission of admission. Critically, Caltech prohibits translating an essay written in another language, which catches out multilingual students who draft in their first language. Common App does not use AI detection tools, because they are unreliable, so enforcement comes from human readers and reported inconsistencies rather than a detector score.
What Common App Says
The Common App fraud policy defines fraud as misrepresenting as your own the substantive content or output of an artificial intelligence platform, technology, or algorithm. A substantiated finding can result in admission revocation or expulsion, and disciplinary steps can include reporting the finding to every member college on your My Colleges list. That last detail is the one students underestimate: a problem found at one school does not stay at one school. Sources: Common App fraud policy; U.S. News.
One widely misunderstood point: Common App does not use AI detection software. It has deliberately avoided those tools because they are unreliable — unreliable enough that OpenAI withdrew its own AI detector. So the mental model of beating a detector is the wrong model entirely. What actually happens is human readers noticing that an essay does not sound like the rest of an application, or an inconsistency getting reported. You are not being scanned; you are being read.
What Yale Permits, in Its Own Words
Yale publishes a dedicated admissions AI policy, and it is more permissive than students often fear. It uses the same fraud language about substantive content or output, and says that using text-generating software to compose personal statements or other written responses may result in admission revocation or expulsion. But it explicitly permits two things: using an AI platform to review your grammar or spelling, and seeking general advice or topic suggestions at the start of the writing process. Yale frames the reason clearly — admissions officers are trying to understand an individual's unique perspective, background, and insights, and a tool that does not know your life cannot supply those. Source: Yale Undergraduate Admissions.
Caltech Publishes the Clearest List — and the Biggest Trap
Caltech's guidelines are the most explicit of the major published policies, and worth reading in full if you are applying anywhere selective, because they make the boundary concrete. Prohibited: copying and pasting directly from an AI generator; relying on AI generated content to outline or draft an essay; replacing your unique voice and tone with AI generated content; and translating an essay written in another language. Permitted: using AI tools like Grammarly or Microsoft Editor to review grammar and spelling; generating questions or exercises to help kick start the brainstorming process; and using AI to research the college application process. Source: Caltech admissions.
Read that prohibited list again and notice the fourth item. Translating an essay written in another language is banned. For a multilingual student, drafting in the language you think in and then translating is an entirely natural workflow, and it feels nothing like cheating — you wrote every idea yourself. Caltech treats it as prohibited anyway, because the English voice on the page would be the model's rather than yours, and voice is precisely what is being assessed. That is the single most likely way a well-intentioned applicant crosses a line this year.
Caltech also offers the most useful test anyone in admissions has published: ask whether it would be ethical to have a trusted adult perform the same task you are asking of ChatGPT. A teacher pointing out a spelling mistake is fine. A teacher suggesting you write about your grandmother is fine. A teacher writing the paragraph for you is not. That heuristic resolves nearly every ambiguous case without needing to memorise anyone's policy. The stated consequence, for the record: failure to comply may result in the rescission of your admission to Caltech.
| Use of AI | Yale | Caltech |
|---|---|---|
| Grammar and spelling review | Permitted | Permitted (names Grammarly, Microsoft Editor) |
| Topic suggestions at the start | Permitted | Permitted as brainstorming questions |
| Researching the application process | Not specified | Permitted |
| AI outlining or drafting the essay | Prohibited | Prohibited |
| Copy-pasting AI output | Prohibited | Prohibited |
| Translating an essay you wrote | Not specified | Prohibited |
| Stated consequence | Revocation or expulsion | Rescission of admission |
The Rule That Actually Keeps You Safe
Since policies differ and not every school publishes one, the practical approach is to comply with the strictest published standard you can find rather than the most permissive one that happens to apply to your favourite school. If your workflow would pass Caltech's list, it passes essentially everywhere. If it only passes at the most lenient school on your list, you have a problem the moment you add a school.
- Safe everywhere: asking AI to generate brainstorming questions about your own experiences, then answering them yourself in your own writing.
- Safe everywhere: running a finished draft you wrote through a grammar and spelling check.
- Safe everywhere: asking AI to explain what a prompt is really asking for, or how the application process works.
- Not safe: asking for an outline and writing to it. Caltech names relying on AI generated content to outline as prohibited.
- Not safe: writing in your first language and translating with AI. Caltech prohibits this explicitly, and it is the trap most likely to catch a sincere applicant.
- Not safe: asking AI to improve the style or flow of a paragraph. That replaces your voice and tone, which is exactly what is being evaluated.
- The test to remember: would it be acceptable if a trusted adult did this for you? If not, it is not acceptable from a model either.
Why the Voice Argument Is Not Just Rule-Following
There is a practical reason these policies focus on voice rather than on effort. An admissions essay is one of very few places where the writing itself is the evidence. Grades show what you did, recommendations show how others saw you, and the essay is supposed to show how you think. A model produces competent, fluent, faintly generic prose — which is precisely the register that makes an essay forgettable in a pile of thousands. Even setting the rules aside, the AI-smoothed version of your essay is usually the weaker one, because the specific, slightly awkward, unmistakably yours detail is exactly what a language model averages away.
Where AI genuinely helps in this process is everywhere except the essay itself: understanding what a prompt is really asking, building a realistic school list, decoding financial aid documents, preparing for interviews, and keeping deadlines straight. Those are legitimate, permitted, and genuinely time-consuming. If you want to compare how different models handle that kind of planning work, LumiChats keeps many current models under one login at a pay-per-day price — useful for the research around an application, and emphatically not for writing the essay, which has to be yours.
01Can I use AI at all on my college application?
Yes, within limits that vary by school. Yale permits grammar and spelling review plus general advice or topic suggestions at the start. Caltech permits grammar and spelling tools, brainstorming questions, and researching the application process. What is prohibited everywhere is AI producing the substantive content of your essay.
02What happens if a school thinks I used AI to write my essay?
Common App treats misrepresenting AI-generated substantive content as your own as application fraud, with consequences up to admission revocation or expulsion, and findings can be reported to every member college on your My Colleges list. Caltech states that failure to comply with its AI guidelines may result in rescission of admission.
03Do colleges use AI detectors on essays?
Common App does not, because detection tools are unreliable — OpenAI withdrew its own detector for that reason. Enforcement comes from human readers noticing that an essay does not match the rest of an application, and from reported inconsistencies, not from a detector score.
04Can I write my essay in my native language and translate it with AI?
Not for Caltech, which explicitly prohibits translating an essay written in another language. This is the rule most likely to catch a sincere multilingual applicant, because the workflow feels legitimate. Assume it is prohibited unless a specific school says otherwise in writing.
05Is using Grammarly allowed on a college essay?
Caltech names Grammarly and Microsoft Editor as permitted for reviewing grammar and spelling, and Yale permits grammar and spelling review. Correcting mechanics is accepted; rewriting sentences for style or flow crosses into replacing your voice, which is prohibited.
06How do I know what my specific school allows?
Check that school's admissions site for a published AI policy — Yale and Caltech both have dedicated pages. Many schools publish nothing, so the safe approach is to comply with the strictest policy on your list. A workflow that passes Caltech's guidelines passes essentially everywhere.
The bottom line: use AI to understand the process, to ask yourself better questions, and to catch a typo in a draft you wrote. Do not use it to outline, draft, restyle, or translate the essay. The stakes are not a lower score but a revoked admission, the schools disagree enough that only the strictest standard is safe, and the honest truth is that the AI-polished version of your essay would probably have been the less memorable one anyway. Write the awkward, specific, true thing yourself.
