When AI makes an essay better but its author is admitted less often, is that lower ability or just distrust?

This explores why applicants who use AI on their essays get admitted less often, even when AI improves the writing: is the penalty catching weaker candidates, or is it readers distrusting AI-assisted work no matter who wrote it? The short answer is that the collection can't fully settle this. What it does show is that the either/or framing is too tidy. When writing becomes cheap to produce, "distrust" and "lower ability" start to blur together.

Start with the central puzzle. In about 7,500 applications to a public policy master's program, most 2025 applicants used AI on their essays despite an explicit ban. Those applicants were admitted at lower rates than comparable non-users, even though AI made their essays better Does AI essay use hurt admissions chances despite quality gains?. That rules out the simplest story, that AI essays are just worse. A separate experiment found that admissions officers can often spot AI writing and rate suspected AI essays lower Do admissions officers penalize essays they suspect are AI-written?. Note the authors' own caveat: the link between that suspicion and the real-world penalty is proposed, not directly measured. Two other things also travel with AI use here: the essay itself and the act of breaking the program's rule.

Here is the part you might not expect: distrust can be a rational response to losing information, not just prejudice. A Freelancer.com simulation shows what happens when written applications stop telling employers anything. Top-quintile workers get hired 19% less often and bottom-quintile workers 14% more often, because effortful writing used to separate able workers from the rest Does cheap writing weaken hiring based on worker ability?. Seen this way, an admissions officer who discounts a polished AI essay isn't necessarily judging the applicant's ability. They're noticing that the essay no longer shows that ability. Hiring shows a similar escalation: applicants send more applications and use prompt injections, recruiters spend more time filtering, and suspicion becomes the default Are job applicants and employers locked in an escalating AI arms race?.

Other evidence points toward bias that isn't about ability at all. AI assistance changes how readers see the writer across all 29 traits one study measured, making them seem more extreme, more confident and more privileged Does AI writing assistance change how readers perceive the writer?. So readers may be reacting to a persona the tool created rather than to the applicant. Accusations of AI use also land on human writers whose text has none of the features that mark AI writing. That suggests suspicion works as gatekeeping rather than detection Do unfounded AI accusations harm human writers instead?. Because AI reliance varies by culture, with Indian writers accepting more AI suggestions than American writers Is higher AI use by Indian writers a confound to control?, a distrust penalty could fall unevenly across groups. Size matters too. When AI help is openly disclosed on an identical article, the penalty is real but small, under 0.15 points on a 7-point scale Does disclosing AI assistance make readers trust articles less?. The much larger admissions effect may therefore come from suspected rule-breaking rather than from AI use itself.

What's missing: no study here measures whether AI users actually have lower underlying ability, for example by tracking how admitted AI users perform later. Nor do we know whether the penalty fades as AI becomes ordinary. The best evidence suggests that people who use AI are expected to be judged less competent, and AI's agent-like nature could keep that judgment alive even after the novelty wears off Does the social penalty for AI use fade as the tool becomes ordinary?. For now, the defensible reading is that the penalty is mostly about trust and signals. It may be partly justified, since the essay no longer reveals much, and partly misplaced, since readers punish a distorted persona or wrongly accuse human writers.