Can students use AI to brainstorm college essays or proofread their applications? Different colleges have different rules about what’s allowed, making it tricky for students and counselors to know where to draw the line.
AI use among high school students is growing quickly, making its effect on the college admissions landscape different from our first review of its impact in 2025. A recent EAB survey found 46% of high school students reported using AI tools in their college search, a jump from 26% in spring 2025.
Colleges continue to update their guidance, too. With early application deadlines fast approaching, we review current policies governing AI use in admissions at the top U.S. universities and liberal arts colleges, complete with analysis on what stood out to us when reviewing them.
We reviewed 32 schools: the top 20 national universities in the U.S. News & World Report rankings (22 including ties) and the top 10 liberal arts colleges. The rankings reflect U.S. News data accessed on October 7, 2026. U.S. service academies were excluded.
Each entry draws on the guidance a school has published on its own admissions site, along with public comments from admissions leadership where they exist.
Policies reflect each school’s published guidance as of October 2026. Admissions offices revise these statements often, so check a school’s own admissions website for current wording before applying in future cycles.
Of the 32 top schools reviewed for this guide, 12 have published no AI policy of their own. However, students should not mistake that silence for permission; nearly all of those schools receive applications through the Common App, whose Fraud Policy bars applicants from submitting the “substantive content or output” of an artificial intelligence platform.
For students, parents, and counselors trying to work out what is acceptable, the picture is a patchwork. A few schools itemize allowed and prohibited uses, several draw a broad line between help and authorship, and others say nothing at all. The groupings below make those differences easy to see.
In August 2023, Common App amended its Fraud Policy to address generative AI, which all applicants agree to when they submit an application through the service. The policy bars applicants from submitting the “substantive content or output” of an AI platform. Thus, the Fraud Policy acts as an AI policy baseline, however vague it may seem, even when a (Common App member) school has published nothing of its own.
What this means: every Common App member school prohibits submitting the “substantive content or output” of an AI platform. Some schools leave it at that; others spell out what this standard means to them.
Two important cases sit outside Common App:
As a brief aside, longtime Common App holdout Georgetown now accepts applications through the platform, which we mention in our guide to 2027 application cycle updates.
The sections below break down each school’s published guidance and public advice, grouped as in the table above.
At Amherst, Harvey Mudd, Pomona, Stanford, UChicago, Vassar, Wellesley, and Williams, our research found no school-specific AI policy and no published admissions guidance on AI use. However, Common App applicants to these schools are still bound by the platform’s Fraud Policy against submitting the “substantive content or output” of an AI service.
Dartmouth, Duke, MIT, and Princeton have no formal AI admissions policy, but leaders at each school have informally commented on best practices.
These schools affirm that submitted work must be the applicant’s own but do not list permitted or prohibited uses.
These schools limit generative AI tools to spelling, grammar, or readability checks.
These schools allow AI to support brainstorming or feedback, but the applicant must write the essay.
These schools list specific examples of allowed and prohibited uses.
This grid pulls the specific uses named in each school’s guidance into one view. A school that is not listed in a row may simply not address that use.
“Allowed” does not mean “encouraged.” Many schools take a nuanced approach to their AI policies. For example, Yale, University of California schools, and Carnegie Mellon all allow limited AI use, but caution applicants against leaning on it to substitute their own experiences.
Many schools allow AI as a partner, but not an author. Eighteen of the 32 schools explicitly permit some AI use, most often proofreading, brainstorming, or research about colleges. Washington University in St. Louis is a notable exception on brainstorming, listing “brainstorming essay topics” among its unethical uses. Caltech offers a useful litmus test: “would it be ethical to have a trusted adult perform the same task you are asking of AI?”
“Proofreading” does not mean the same thing everywhere. Carnegie Mellon counts suggested structural and vocabulary edits among its approved uses, and the University of California permits help with “readability.” Brown, by contrast, limits AI powered tools to only basic spelling and grammar review, Bowdoin and Swarthmore bar changing an essay’s tone, and Washington University in St. Louis lists grammar and spelling review as “sometimes” acceptable, allowing the assistance only when a teacher or counselor could also give it ethically.
Duke views AI (and essays) distinctly. According to the Duke Chronicle, Duke University does not score essays and reads them to assess character rather than prose quality, allowing the school to take a more dynamic stance on applicant AI use. As Duke’s Dean of Admissions puts it, “We don’t think of [AI] as an inherently bad tool for student to use,” reasoning that it “can only write about what the students put into it.”
This contrasts with schools that insist an essay reflect only the student’s own ability. For example, Harvard requires that submissions “reflect [the applicant’s] own content, writing style, and English proficiency” and treats fraudulently created essays as a violation of its Honor Code. Whether more schools follow either approach is worth watching.
Where schools mention translation, they prohibit it. Each school that mentions using AI to translate materials prohibits the activity, including translating an essay from another language into English. Thus, it would be advisable for applicants to not assume it is acceptable at a school that is silent on the question, an important insight for international students who find written English challenging.
AI policies can reach beyond the personal statement. Notre Dame requires that “essays, short answer questions, activities, and any other material submitted” be the applicant’s own work, and Cornell and Washington University in St. Louis bar using AI to create portfolio work. Northwestern goes a step further, listing AI reworking of an essay written for a different school as a problematic use. Families who equate an “AI policy” with the personal statement may be missing how much of the application it covers.
Many top liberal arts colleges have no published AI admissions policy. Counselors and families should still look up each school’s guidance before drafting, but finding nothing is not unusual, particularly at liberal arts colleges: six of the 10 reviewed have no school-specific policy. A published policy is usually easy to find through an online search. Where none exists, the Common App’s Fraud Policy (if applicable) and any public comments from admissions leaders are the clearest reference points.
Across these 32 schools, policies differ in length and detail but converge on one idea: an applicant should have ownership over the thinking and the words in an application. Where it is permitted, AI can stand beside a student’s writing as a teacher or counselor might, but it cannot stand in for the student’s voice.
Because guidance is still taking shape, the most dependable approach is to read each school’s published expectations before drafting, and to treat any gap as a reason for caution rather than permission.
In limited ways, depending on the school. Eighteen of the 32 schools we reviewed explicitly permit some AI use, most often proofreading, brainstorming, or researching colleges. Schools that itemize uses, including Bowdoin, Caltech, Cornell, and Northwestern, prohibit drafting or generating essay text. Always check each school’s guidance before applying. There are no cases among these schools where directly submitting AI output without disclosure is allowed.
Regardless of whether it is allowed, authenticity is at the heart of many admissions essays, especially short answer responses and the longer, deeply personal stories. Even when one can use an AI model, schools generally advise their applicants to develop their authenticity independently and discourage heavy AI use.
It depends on what “revise” means. Carnegie Mellon allows suggested structural and vocabulary edits, and the University of California permits help with readability. Brown limits AI to fixing basic grammatical and spelling errors, while Bowdoin and Swarthmore bar using AI to rewrite portions of an essay or change its tone. Make sure to check each school’s guidance before applying, or ask the admissions office directly.
Common App amended its Fraud Policy in August 2023 to bar applicants from submitting the “substantive content or output” of an AI platform. Most highly selective schools are Common App members, so this works as a baseline even where a school has published nothing of its own.
Twelve of the 32 schools we reviewed have no school-specific policy, including Amherst, Duke, MIT, Princeton, and Stanford. However, the absence of a policy rarely means no guidance exists. Nearly all of these schools use the Common App, and some have commented informally on AI use.
MIT was the only school researched that does not belong to Common App and does not have a specific policy of its own. Despite this, senior admissions officials at MIT have publicly commented against the use of AI generated writing.
Not safely. Every school that explicitly mentions AI translation prohibits it (Bowdoin, Caltech, Cornell, Northwestern, and Swarthmore). Applicants shouldn’t assume it’s acceptable at a school that is silent on the question, which matters most for international students.
Schools tend to focus on whether a student’s essay is the applicant’s own, and admissions teams have a few ways to test that.
The Common App’s Fraud Policy bars AI-generated content or submitting plagiarized essays, and Penn offers a litmus test: if you couldn’t explain how you arrived at what you wrote, in an interview or conversation, it isn’t ready to submit. Colleges may also ask applicants to verify that content or prose in their application is their own if something appears inconsistent.
Admissions officers may use AI detection tools, but these are usually discouraged as a stand-alone indicator of AI generated writing due to their tendencies to mark false positives. Many admissions offices require some measure of human judgement or human review before ruling on application fraud.
Doing research is commonplace. How do you choose the research opportunity that makes a difference?
Join us for a free online info session to learn about Pioneer
Check exclusive sharings From directors of prestigious programs
Sign up to receive alerts for upcoming deadlines and other opportunities for high-achieving students.
Thank you for your interest in Pioneer’s Global Problem-Solving Institute (GPSI). The application for the Fall 2025 terms is now available. Please select your country/region below:
