Essays
Navigating AI in college essays — what admissions readers actually detect
Generative AI has changed what counts as a strong essay. Here's what admissions officers can now detect, what they can't, and how to use these tools without compromising your voice.
Hao Cui · June 10, 2026 · 1 min read
Most families assume admissions officers are still reading essays the way they did in 2019. They are not. The introduction of ChatGPT to the cycle changed three things at once — and the families who win in this environment understand all three.
What readers can detect
Admissions readers are humans, but they read thousands of essays. They have developed an ear for what AI-generated prose sounds like before any detection tool flags it: balanced clause structures, generic emotional vocabulary, and conclusions that resolve too tidily.
The signature of AI is not bad writing. It is average writing dressed in formal clothes.
What they can't detect
A student using AI as a research assistant — asking it to summarise scholarship on a topic, to challenge an argument, or to generate counter-positions — leaves no trace. A student using AI to write the essay leaves a strong one.
The Oak rule
We tell our students they may use AI freely for:
- Brainstorming and idea generation
- Researching the texture of a topic
- Pressure-testing an argument
- Identifying weak transitions
They may not use AI to:
- Generate prose that ends up in the essay
- "Polish" or "tighten" their writing
- Translate notes into paragraphs
The first set sharpens thinking. The second set replaces it — and admissions officers can hear the difference.
What this means for parents
The student your child becomes in this process matters more than the essay they submit. AI accelerates research and weakens writing. Use it as an accelerator. Never as a substitute.