Colleges Race to Detect AI in Admissions Essays

Admissions offices are confronting a genuinely difficult authenticity problem this application season, one that's reshaping both sides of the college admissions process simultaneously. As AI tools have become ubiquitous, a review of tens of thousands of admissions essays at a selective college found a measurable increase in homogeneous language after generative AI tools became widely available in 2022, according to Inside Higher Ed's reporting on a study by researchers at Cornell and Carnegie Mellon Universities.

The scale of AI use in college essays is already substantial. According to a 2024 survey cited in the same research, about half of college applicants use AI to brainstorm their essays, and one in five use it to generate a full first draft.

One of the study's more counterintuitive findings involves who's actually turning to AI most. The researchers suggest high-income students may be less likely to use AI in their essays specifically because they have greater access to other writing support, like private consultants or tutors, according to Inside Higher Ed's reporting. That finding pushes back against a common assumption that AI essay assistance primarily benefits wealthier applicants, and the study's authors argue admissions offices should consider these wealth disparities when evaluating essays going forward.

AJ Alvero, a Cornell sociology professor and study co-author, raised a deeper concern about what's actually being lost. "The essay is designed to give applicants an opportunity to highlight the idiosyncrasies of their life," Alvero said, and the growing homogenization of essay language threatens the essay's core purpose as a genuine window into an applicant's individual voice.

A separate Cornell study found that current AI models genuinely struggle to produce essays that sound authentically like a specific applicant, even when directly prompted to try. "Tools like ChatGPT can give solid feedback on writing and are likely a good idea for weak writers. But asking for a full draft will yield a generic essay that just does not sound like any real applicant," said Rene Kizilcec, the study's senior author, according to Cornell's own reporting on the findings. The research specifically found large language models struggle with "steerability," meaning they can't reliably adapt their writing style to match an individual applicant's authentic voice, making them a genuinely poor choice for writing with this much at stake.

The response from admissions offices has moved beyond simply banning AI use. This fall, Virginia Tech is debuting an AI-powered essay reader specifically designed to speed up application review, expecting to inform students of admissions decisions a full month sooner because of the tool's help sorting tens of thousands of applications, according to earlier reporting from VPM. Caltech is launching its own AI tool this fall specifically designed to look for "authenticity" in students who submit research projects alongside their applications, according to Caltech admissions director Ashley Pallie, a genuinely notable use of AI to detect other AI, or simply overly polished, inauthentic work more broadly. This dynamic mirrors the broader pattern we've tracked in higher education, including our coverage of Canadian universities grappling with the same AI authenticity crisis and the Brown University AI cheating scandal earlier this year.

This story is worth understanding for any business in education technology, admissions consulting, or hiring, given the same underlying authenticity problem colleges are grappling with applies directly to job application materials, cover letters, and interview preparation. If AI genuinely struggles to authentically replicate an individual's voice at scale, that's a meaningful signal for HR teams currently worried about AI-generated job applications flooding their pipelines.

For businesses building AI detection or authenticity verification tools, the college admissions space is becoming a genuine proving ground worth watching closely, since the stakes, methods, and countermeasures being developed here will likely translate directly into hiring and other high-stakes evaluation contexts.

College admissions offices are grappling with growing homogeneity in applicant essays since AI tools became widely available in 2022, according to a Cornell and Carnegie Mellon study, even as separate research shows current AI models genuinely struggle to authentically replicate an individual applicant's voice. Roughly half of college applicants now use AI to brainstorm essays, with one in five using it for a full first draft. Schools including Virginia Tech and Caltech are deploying their own AI tools to speed up application review and screen for authenticity, respectively.