Essays that were likely AI-written got admitted less often even when they read better, so what is an essay for?

This explores why AI-assisted essays can read better on paper yet still hurt an applicant's chances, and what that suggests about what readers are really evaluating. The starting point is a study of about 7,500 applications to a public policy master's program. In 2025, a majority of applicants submitted essays that were likely written mostly by AI, even though the program explicitly banned it. Those applicants were admitted at lower rates than comparable applicants without detected AI use, even though AI raised essay quality Does AI essay use hurt admissions chances despite quality gains?. A companion experiment found that admissions officers can often tell AI writing from human writing and rate suspected-AI essays lower Do admissions officers penalize essays they suspect are AI-written?. One caveat: the authors propose this detection effect as an explanation for the penalty but did not measure it directly. And because the program banned AI, part of the penalty may simply be for breaking the rule.

The deeper answer may be that the essay was never really graded on quality. It was graded as a signal. Research on Freelancer.com makes this clear. When writing becomes cheap, employers lose the effortful signal that once separated able workers from the rest. In a simulation that removed written signals, top-quintile workers got hired 19% less often and bottom-quintile workers 14% more often Does cheap writing weaken hiring based on worker ability?. From that angle, a polished AI essay is a bit like a forged credential. It may be better prose, but it tells the reader less about you. Researchers reviewing evaluation practice make the same point: rhetorical polish shouldn't be read as merit, because evaluators often mistake AI text for human writing and rate it higher Does polished writing actually signal better quality work?. Admissions readers who learn to spot AI may be correcting for that bias, and sometimes overcorrecting.

AI also changes who seems to be speaking. A study of nearly 3,000 writers and 11,000 readers found that AI assistance shifted all 29 measured traits of how readers perceived the writer. Writers came across as more confident, more extreme, more agreeable and more privileged Does AI writing assistance change how readers perceive the writer?. Writers rarely push back on this. They edit AI paragraphs only 23% of the time, and their edits leave the text about 96% unchanged Do writers actually edit AI-generated text before publishing?. Some of this is cultural. AI suggestions pull Indian writers toward Western phrasing and references Do AI writing assistants push non-Western writers toward Western styles?. In admissions, the essay is the one place an applicant's own voice is supposed to come through, so this kind of flattening defeats its purpose.

There is also a texture readers seem to pick up on. LLMs master grammar and structure but avoid taking evaluative positions, which produces prose that is organized but argues for nothing Why does AI writing sound generic despite being grammatically correct?. In online discussions, AI writing tools raised participation but made comments seem more generic and less authentic. That perception even spread to conversations among people who hadn't used the tools Do AI writing tools improve online discussion or degrade it?. The penalty for simply knowing AI was involved is real but small: disclosure lowered ratings of an identical news article by less than 0.15 points on a 7-point scale Does disclosing AI assistance make readers trust articles less?. That suggests admissions penalties are less about bias against AI as a label and more about what readers think the essay no longer reveals.

The twist: if readers judge essays by signal and authenticity rather than polish, banning AI may not be the lever that matters. In a randomized trial at ICML 2026, banning LLM use in peer review rather than allowing limited use barely changed scores or decisions, and many reviewers broke whichever rule they were given Does banning LLM use in peer review change review outcomes?. The corpus's real question for admissions may be what an essay can still prove once anyone can produce a good one.