Caltech Class of 2028: How to Answer the AI-in-Admissions Question
加州理工Class 2028关于是否应用AI申请怎么回答?
I was probably among the earliest people in this industry using AI for admissions writing — fine-tuning on collected essays from GPT-2 onward, all the way through GPT-4 and Claude. Now I watch agencies publicly declare they write "100% by hand, zero AI." It has the energy of a handloom protest during the Industrial Revolution.
My original motivation was twofold. First: the tutoring industry is full of bad actors. Students who get burned by unscrupulous agencies are countless, and the college application process should not be this commercial. A student going abroad who cannot research a school, fill in a form, or write about their own life is, frankly, taking a risk. Second: for certain things — particularly graduate-level personal statements in technical fields — AI genuinely produces better output than a human tutor without the relevant background knowledge.
So it was only a matter of time before universities responded. Caltech became the first to formally ask Class of 2028 applicants how they used AI.
What Caltech said you cannot do (plain language)
- Generate text and paste it in directly.
- Let AI produce your structural outline or brainstorm your ideas for you.
- Let AI transform your voice into Rousseau, Thoreau, or J. K. Rowling.
- Write in Chinese and have AI translate it into English.
What Caltech said you can do
- Grammar and spelling checks.
- Use AI to ask you questions — brainstorming and self-discovery.
- Use AI to learn about the application process at a school.
Matching against my own practice: of the prohibited uses, I had previously done 1 and 3. Of the permitted uses, I do all three (currently focused on 3).
The key supplemental principle: adult equivalence
Caltech added a clarifying passage that is actually the most useful part: if the help AI provides is equivalent to help a human adult could give you, then whether it is acceptable depends on whether that human help would have been acceptable. A teacher checking your grammar and spelling — fine; AI doing the same — fine. A consultant writing your essay for you — not fine; AI doing the same — also not fine.
Three conclusions follow:
- Agencies that ghostwrite essays for students were already doing something unethical. AI doing the same makes it no different.
- Agencies loudly advertising "100% handwritten, no AI" are being dishonest in both directions.
- Caltech is actively endorsing the legitimate use of AI in the application process — the same logic as Oppenheimer and nuclear physics: the technology is neutral; the use determines the ethics.
Writing reflects thought. If AI helps you express your thoughts more precisely, use it. If it doesn't, don't.
Caltech also stated clearly that however you answer the AI question, it will not affect your admissions outcome. So: if you're aiming for Caltech, go compete and go do research. Nothing else to see here.