2023 Admissions: The Year AI Entered the Room

Let me start with the numbers. This past cycle I personally managed eight students from start to finish: three US-high-school students (USC, UC Berkeley, NYU), one Bay Area student (NYU), three mainland-China students (NYU ×2, UCLA), and one US-undergrad transfer (Tufts → UChicago). Beyond these eight, other agencies pushed me additional referrals and postgraduate cases — I did not handle their complete applications and many of their offers never reached me, so I will leave those aside.

The headline of the season is simple: ChatGPT arrived in college admissions, and it changed almost everything about how the work gets done — not what the work is. The hours did not shrink. But the texture shifted considerably.

The Notion System

Back in 2022, with too many students and too much moving information, my brain simply stopped being sufficient. So I built a Notion-based management system — primarily for tracking application progress and managing activity lists, with the Common App activity table importable directly. Parents and students get a shared view; the entire process is transparent. That system is still running and will keep improving.

How We Actually Used AI

My honest position on AI in college applications, stated plainly:

1. Submitting without any AI help at all is foolish.
2. Submitting work that is entirely AI-generated is dishonest.

Carnegie Mellon's own AI guidance makes a useful distinction: if the help AI gives you is equivalent to the help you could get from a strong English teacher or a more experienced writer, it is acceptable — grammar correction, language polish, content review. If AI is ghostwriting your activities or essays wholesale, that is no different from hiring someone to do it for you, and no school can read past it.

My workflow in 2023 shifted accordingly. Before AI, we might brainstorm briefly and start drafting. Now we spend much more time in conversation first — mapping the overarching narrative, establishing the arc of the essay, deciding which experiences are load-bearing and how they connect across the application as a whole. If a student was stuck at the start, I would point them to a GPT I developed for Common App essay thinking; they would work through it on their own before our next session.

On the activity list side, I built a separate GPT: feed it a resume or a quick verbal description of what a student did, and it outputs a fully formatted Common App entry — action verb openings, specific numbers, the whole structure. That shortened one of the more tedious parts of the process meaningfully.

The essays that emerged from this AI-assisted process also passed AI detectors without strain. The highest GPTZero score I saw was 45%; most fell under 30% and registered as human-written. This is not a trick — it is the natural result of a genuinely human-led process that used AI as a language tool, not a ghostwriter.

Using AI to express real thoughts with precision is good craft. Expecting AI to bypass the essay entirely is wishful thinking.

A word on the market: in 2023, a large share of agencies started pumping AI-generated essays. Some came to me for revision after the fact — they had a ChatGPT draft and wanted me to fix it. The output was unmistakable. If an agency's draft is full of phrases like "I have always been passionate about" and "this experience taught me the importance of," run it through GPTZero. If it reads machine-generated, it almost certainly is. Master's applicants especially need to be alert to this.

AI humanizer plugins are almost uniformly useless, by the way. The output destroys whatever rhythm the essay had. There is no shortcut through the essay.

Early Decision: The NYU Pattern

Five of my students applied Early Decision to NYU (none to NYU's School of Professional Studies). Four were admitted; the fifth was deferred and later admitted to USC in the regular round. Why did they all end up at NYU? Looking back at each case:

• NYU was a genuine dream school — two students.
• Profile was strong but top-30 uncertain; NYU felt like a winnable bet — two students (one admitted, one deferred).
• Decided to apply ED late, almost on a whim — one student (admitted).

Beyond my eight, I helped other agencies with materials for two more NYU admissions. Even so, after six admits I still cannot tell you with confidence what NYU is really selecting for. My rough theory: high GPA, high standardized scores, clean and organized materials, some affinity for the city, and possibly a sense that a student is going somewhere from nowhere — from somewhere quiet to the largest stage available. But I genuinely cannot pin it down further than that.

Among other ED outcomes: one Duke application was rejected (later admitted to UCLA); one Yale application was rejected (later admitted to Johns Hopkins and UC Berkeley). All in all, a strange and productive year.

I will end with a note on claims. Agencies and counselors who say they know exactly what Harvard wants, or exactly why Stanford passes on a particular student, are — in my view — charlatans. Excellent is excellent in many directions. Universities need diversity. No formula survives contact with a real admissions committee.