How AI Reads Your College Application (and What It Misses)

More than half of U.S. admissions offices in 2024 used AI to do an initial pass on applications. The models are trained on past admit decisions, which means they reproduce past bias — favoring writing patterns common in well-resourced high schools and penalizing AAVE, code-switching, and ESL phrasing.

Some examples

• A 2023 leaked Common App audit showed AI rubrics gave lower 'narrative coherence' scores to essays mentioning food insecurity, even when essays were equally well-written.
• Schools using Slate's AI assistant rank essays partly on vocabulary 'sophistication' — which correlates more with zip code than with ability.
• Stanford's CARTA tool, used in graduate admissions, was paused after admitting it ranked applicants partly on undergraduate institution prestige.
• Some essays now get scored twice: once by AI, once by a human, and only flagged for closer review if the scores diverge by more than 20%.

Try it!

Take any draft essay and run it through a plain readability checker (hemingwayapp.com is free). Aim for grade 8-9, lots of short sentences, real verbs. That's what scores high with both humans and the bots.