Connect with us

AI

The Coaching Trade That Makes College Essays Sound Like AI

A $3.0 billion consulting trade taught the copyable personal-essay voice. ChatGPT just made that voice obvious, and colleges are cutting the short prompts.

Published

on

U.S. admissions consulting booked $3.0 billion in 2025, and the personal essay is one of the products that money buys. When college essays now read like ChatGPT, the cause is often a coaching trade that spent years teaching students to write in a voice machines copy with ease.

Selective colleges are answering by cutting the short extra prompts that used to do extra sorting. The 650-word Common App statement has to carry more of the file, and it is the format both coaches and chatbots already know by heart.

The Consulting Trade That Taught Everyone to Sound Alike

Navagant’s Q4 2025 sector report put admissions consulting revenue in 2025 at $3.0 billion in the United States, with a forecast of $3.2 billion by 2029. The same note said firms that folded in AI for essay review, predictive modeling, and feedback are the ones buyers now treat as category leaders. Private equity led 53 percent of disclosed 2025 deals in the wider prep sector.

The Independent Educational Consultants Association has watched the headcount explode. IECA estimated about 1,500 full-time independent educational consultants in the United States in 2010, about 3,500 in 2015, and 8,500 to 10,000 full-time consultants by the end of 2024, plus about 1,500 working abroad and another 10,000 to 15,000 part-timers.

THE CONSULTING BENCH

Measure Figure
U.S. admissions consulting, 2025 $3.0 billion
Full-time U.S. consultants, 2010 about 1,500
Full-time U.S. consultants, end of 2024 8,500 to 10,000
Typical hourly fee, IECA FAQ just under $140
Mean college package, 2022 IECA survey $6,304

IECA says public-school counselors in some districts carry 600 to 1,000 students, which is how a private shop became the person who actually sits with the essay. It still cites a Lipman Hearne study that found 26 percent of high-achieving students used an independent consultant, up from 3 percent in an earlier survey of the same kind of family.

The association also draws a bright line its members are supposed to honor. IECA forbids writing a student’s essay or editing it heavily. Light review, the way an English teacher might mark a draft, is allowed. The commercial internet around those members never accepted that limit. Blogs, sample files, and paid packages taught a house style: a vivid scene, a tidy hardship, a closing sentence that begins “I learned that.”

Johns Hopkins Still Posts Essays That Worked

Johns Hopkins still publishes essays that worked from Hopkins, with notes from its admissions committee. The page tells applicants to be original. It also hands every junior a public stack of pieces that already got someone in.

Coaches and content sites then recode those samples into moves. Start a club and extract leadership. Volunteer at a hospital and extract compassion. Survive a family shock and extract vulnerability. The students who follow that map are often sincere. The writing still stops at the lesson, as if a life were a prompt with a correct theme word.

THE FORMULA THAT TRAVELS

  • The club arc: A student founds something, hits a snag, then names leadership as the takeaway.
  • The hospital arc: A volunteer shift produces compassion, grit, or a future in medicine.
  • The family arc: A hard season at home is translated into vulnerability or resilience before the student has sat with it.
  • The close: Heightened imagery does the showing, then an abstract noun (growth, passion, identity, authenticity) does the selling.

Before those archives sat in one place, a personal statement was a quirk of the American file, and a lot of it was rawer because nobody had a national answer key. The polished version is easier to admire in a single read. It is also easier to imitate, which is the trait a language model needs.

What Happens When ChatGPT Writes Your Life

Researchers at Cornell’s Bowers College of Computing and Information Science compared 30,000 human college essays with drafts from eight large language models built by OpenAI, Meta, Anthropic, and Mistral. They used essays written in the three years before ChatGPT’s November 2022 debut, then asked the models to answer the same question, including a second pass that named the writer’s race, gender, hometown, and parents’ education.

The models did not become more specific when they were given a real person to imitate. They repeated keywords from the prompt and listed biographical facts in a set order. ChatGPT wrote, “Growing up in Lexington, South Carolina, with my Asian heritage, I often felt like a bridge between two cultures. My parents, both college graduates, emphasized the importance of education and hard work.”

Asked to sound like a particular applicant, it sometimes sounded less human. One unsteered draft opened on an eighth-birthday robot kit. With extra personal data, the same kind of prompt produced, “Growing up in Rabat, Morocco, I have always been acutely aware of the intricate tapestry of my heritage. As a biracial individual with a Black mother and a White father, my identity has been a powerful lens through which I view the world and an integral part of my journey.”

Tools like ChatGPT can give solid feedback on writing and are likely a good idea for weak writers. But asking for a full draft will yield a generic essay that just does not sound like any real applicant.

Rene Kizilcec, associate professor of information science, Cornell Bowers

Co-author AJ Alvero said students need to sound like themselves and only themselves, and that a full AI draft can put them at a disadvantage. The team trained a classifier to separate human essays from model essays and got near-perfect accuracy, which is a warning to anyone who thinks a chatbot can hide in a 650-word statement. A foundry10 report cited in the same Cornell account estimated that about 30 percent of students already use these tools to write essays. First author Jinsook Lee’s advice is blunt: write the first draft yourself, then use a model for polish only if the college allows it.

Selective Colleges Are Dropping the Short Prompts

The extra essays were supposed to be the place a student showed fit. They were also short, patterned, and easy for a model (or a coach with a swipe file) to fill. Several selective campuses have now stripped them for the 2026-27 cycle rather than pretend they still sort anyone.

THE SHORT PROMPTS THAT WENT AWAY

School What applicants no longer write
Tulane The “Why Tulane?” prompt, paused
University of Georgia The “meaningful book” essay
Cornell The university-wide essay; school prompts remain
Washington University in St. Louis An optional essay many students treated as required
University of Virginia A supplemental essay, cut in the prior cycle
Brown “In one sentence, Why Brown?” and “What three words best describe you?”

Tulane spokesman Michael Strecker said concerns about AI-written essays “were definitely a factor” in pausing the why-us prompt. David Graves, executive director of undergraduate admissions at Georgia, said he figures every admissions office now worries about essays written by AI, even though Georgia also framed its cut as a way to simplify the file. Brown’s two ultra-short answers were the kind of prompt a model completes in one pass: a slogan, then three adjectives.

What remains, at most Common App schools, is the long personal statement plus whatever a campus still bothers to ask. That concentrates the old coaching script onto one page. It also concentrates the advantage of a human editor who knows the script is dead.

Why Detectors Keep Flagging Honest Writers

Parents keep asking whether a detector can settle the fight. The tools are bad at the exact students admissions offices say they want more of. Stanford researchers testing seven detectors found they flagged 61 percent of essays by non-native English speakers as AI-written, while writing from U.S. eighth graders mostly passed as human. When ChatGPT rewrote the non-native essays with fancier vocabulary, the false-flag rate fell. Predictable English reads as machine English. That is how a lot of honest second-language writers sound, and it is how a lot of coached native writers sound too.

Essay coaches now describe drafts that carry chatbot tics even when a student insists no model was open. The giveaway is often a mirrored sentence, the “it’s not just this, it’s that” move, or a sudden jump into inflated nouns. Some of those students did paste from a chatbot. Others have spent two years reading chatbot summaries instead of books, then reproduced the only prose they still see. The sameness is a reading diet as much as a cheating method.

WHAT THE SURVEYS SHOW

  • Kaplan field period: 220 admissions officers at national, regional, and liberal-arts colleges, emailed in July and August 2025.
  • Writing the essay: 2 percent of colleges had a policy allowing generative AI to write it, 30 percent banned that use, and 68 percent had no policy.
  • Brainstorming and feedback: 27 percent allowed AI for ideas and 21 percent allowed it for comments on a student-written draft; most campuses had no rule on either.
  • Officer mood: 50 percent unfavorable toward generative AI in essays, 26 percent neutral, 14 percent favorable.

Dewey Smart’s tracker of 50 selective schools, last checked on August 18, 2026, found 26 with some public position on generative AI in essays: 21 allowed limited use such as brainstorming or grammar, two (Harvard and Brown) explicitly barred AI-generated essay content, three offered guidance only, and 24 had no official statement on their admissions pages. A ban without a working test is a poster. A missing policy is a shrug.

The Common App Fraud Rule Nobody Reads

The floor that every Common App applicant already signed sits in a fraud policy last updated in August 2023. It treats as fraud the act of submitting plagiarized writing, or of presenting as original work another person’s thoughts, or the AI platform output as original work. The exact phrase is “the substantive content or output of an artificial intelligence platform, technology, or algorithm.”

Yale’s undergraduate admissions office repeats that clause and adds teeth. Submitting personal statements “composed by text-generating software may result in admission revocation or expulsion.” Grammar and spelling checks, and general topic suggestions at the start, do not count as fraud there. Notre Dame draws a similar line: brainstorming and spelling are fine; drafted or substantially written AI content is not.

The word that does the work is “substantive,” and Common App has not published a test for it. Kaplan senior vice president Jason Bedford called the campus posture laissez-faire almost three years after ChatGPT’s debut, in the firm’s November 3, 2025 release, and said the gray zone pushes some students to overuse the tools while others avoid them out of fear.

Some students may overuse AI, while others may avoid it entirely for fear of crossing an unseen line. We expect to see more schools implement clear policies or guidance in the near future.

Jason Bedford, senior vice president, Kaplan

Common App does not run a detector across the platform. Member colleges decide whether to look, and most still do not say how. A fraud finding can follow a student to every campus on the list. That is a severe penalty attached to a rule most seniors never read, enforced by offices that still cannot tell a coached human from a steered model.

Consultants Now Sell a Human-Sounding Draft

The same industry that taught the copyable voice is now productizing the opposite. Coaches sell “anti-formula” sessions: notice a boring habit, stay inside one scene, refuse the theme-word ending. Platforms sell the cheap version of that help. U.S. News & World Report acquired Sups AI, a tool that brainstorms essay ideas, organizes applications, and returns structured edit notes, which is the human consultant’s workflow with a model in the middle.

WHAT THE NEW PACKAGE INCLUDES

  • Allowed machine help: brainstorming and grammar, the uses Yale and many peers already bless.
  • Forbidden machine help: a full draft passed off as the student’s, which is fraud on the Common App and a revocation risk at Yale.
  • Paid human help: an editor who now charges for making a draft sound less like the sample files the shop used to circulate.
  • The leftover signal: a 650-word statement that still rewards families who can buy taste, time, and a reader who will kill the “I learned that” line.

IECA’s ethics code still says the student has to own the sentences. The market around that code sells a smoother path through a process school counselors cannot staff. ChatGPT did not invent the interchangeable essay. It made the house style impossible to ignore, then offered to write it at no cost, which is why the paid shops now advertise a human voice as the premium add-on.

Colleges can keep asking 17-year-olds to perform self-knowledge on a deadline. They can also stop scoring the students who do it in the cadence those shops already sold. The file still contains one long statement. The people who get a distinctive one are, more than ever, the people who can pay for a reader who knows the old cadence is now a tell.

Harry is the editor of Oton Technology, an independent site he owns and edits, covering the part of technology that people actually have to act on. After ten years in journalism, first reporting and then editing, he works from primary material by habit: the advisory rather than the write up of it, the filing rather than the press release, the changelog rather than the launch video. Every figure in an article carries its source and its date, and where a number comes from a vendor or an analyst model rather than a count, he says so plainly instead of letting it stand as established fact. What he leaves out is anything he could not verify himself, which on a beat full of unnamed supply chain claims removes a great deal. That standard applies across all the sections the site publishes for an international audience, from artificial intelligence and security to phones, computers, gaming, crypto and the software businesses depend on. He corrects errors in the open and labels them, because a site that hides its mistakes is asking readers to trust the rest on nothing.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending