What generic AI gets wrong with this prompt
An essay about your optimal experience at a specific school is the easiest prompt in the world to answer with someone else’s material. That’s precisely how generic AI answers it.
It assembles a brochure collage
Cluster visits, named clubs, the New York advantage — generic AI builds the essay from Columbia marketing material because that’s what it has. The draft describes what Columbia offers, which the committee wrote. Worse, models routinely misname clubs and classes or cite ones that no longer exist, and one wrong name undoes the whole “specific” instruction.
Ask your AI — “For every Columbia resource in my draft, tell me how I tied that to my growth, goals or contributions to Columbia”
It answers “attend” when the verb is “co-create”
Generic AI writes an itinerary — I will join, I will take, I will attend — because involvement lists dominate its training data. Columbia chose a stranger verb. Co-create means the environment is different because you were in it: something you’d start, host, organize, or change. A consumption plan answers a question the school deliberately didn’t ask.
Ask your AI — “Which items in my draft would exist at Columbia whether or not I enrolled? What in this essay only happens if I’m there?”
The experience has no through-line
Ask for an optimal experience and generic AI enumerates — a club here, a class there, a trek — with no logic connecting the choices. Models list; they don’t compose. But the committee reads this essay to learn what kind of classmate you’d be, and a grab-bag of activities describes a schedule, not a person.
Ask your AI — “If you had to name the single theme connecting every choice in my draft, what is it? Which items don’t fit it?”
It hands the prompt’s own words back as an answer
“I will foster a sense of belonging, agency, and partnership” — generic AI mirrors the vocabulary of the question because echoing the prompt is the cheapest way to sound aligned. The committee wrote those words; reading them back adds nothing. The essay works when your specifics make the prompt’s abstractions concrete, not when they restate them.
Ask your AI — “Which phrases in my draft also appear in the prompt itself? For each one, what specific plan of mine was it standing in for?”