A false positive is when a detector flags genuinely human writing as AI. The two groups it hits hardest are polished essayists and non-native English writers — both for the same reason: formal, fluent prose can share surface features with model output even when no model was involved.
Personal statements are a high-risk genre
Personal statements are usually drafted, rewritten, drafted again. By the seventh pass the rhythm has been smoothed, the vocabulary has been levelled up, and the writer has been told to "sound professional". That sanded surface is exactly what a statistical detector reads as AI — even though every word came from a human writer reaching for the right one.
What to do if you've been flagged
- Don't panic. A score is not a verdict — most institutions know this. Re-read the policy your institution actually published, not the rumour.
- Look at the per-sentence breakdown. Are the flagged sentences the ones you most heavily edited? They probably are. That's drift toward the average, not evidence of AI.
- Put some burstiness back. A few short sentences next to your long ones lifts the human signal almost immediately.
- Keep your drafts and your version history. If you're ever asked, the timestamped evolution of your prose is a much better defence than any score.
Before
Throughout my academic journey, I have consistently demonstrated a profound passion for biomedical research, which has been instrumental in shaping my aspirations and motivating my pursuit of scientific excellence.
After
I learned what biomedical research actually felt like in the second year of my undergraduate, in a lab basement in Manchester, pipetting cell lysates at 11pm with the same playlist on repeat. I'd thought it would feel glamorous. It didn't, and that turned out to be the point.