Which tier of university does your essay reach?

Researchers have a name for AI's habit of pleasing users: sycophancy — models are trained to prioritize what you want to hear. Anthropic's experiment is blunt: the same piece of writing gets noticeably warmer feedback when introduced as “I wrote this” than as “a friend wrote this.” A generic AI's praise is, in large part, for the person asking — not for the essay. Our rubric wasn't improvised from experience: it is distilled from more than 110,000 successful essays across every tier. Held to that rubric, the AI cannot distort in either direction — upward, even praise must point to the exact sentence in your essay; downward, a dimension with no significant weaknesses says so plainly: “nothing worth editing in this dimension.” Invented problems are worse than missed ones: they spend your revision budget where nothing needs fixing.

No “the structure could be tighter” style of empty truisms — every comment maps to a specific sentence you can check anytime.

The risk of AI tells isn't rumor — it's written policy: the Common App treats submitting the substantive content or output of an AI platform as application fraud; Brown and Georgetown restrict AI use in applications; Yale states that admission can be rescinded over it. And beyond policy, admissions officers who read thousands of essays each season develop an instinct for the “AI voice” that arrives earlier than any detector — the same sentence patterns, the polished but soulless structures, recognizable at a glance. Rather than submitting and hoping, know where you stand first: the evaluation examines your essay's language texture, and a pronounced AI writing style is flagged in the report and reflected in the score. This is a signal, not a verdict — whether and how to revise remains your call.

Detectors produce false positives; policies don't bend — “reads like a human wrote it” is the only answer that's safe on both sides.