Name the tool and say what you used it for. Then say what stayed yours, including what it didn’t do. Most schools’ AI disclosure formats fit in three sentences, and if you’re struggling with the assignment, you’re usually struggling with something else. You haven’t decided what you actually want to admit to.
“How you describe your AI use is itself evidence of your judgment.”
— Yvette Romero, HBS MBA and founder of MBA.AI
Which schools require AI disclosure and in what format?
HBS asks a yes/no question. Did you use AI in completing the application? If yes, you get a 75-word statement to describe how and where.
Kellogg wants a citation at the end of the essay itself, formatted as (Name of Tool, URL), the same way you’d cite any other source.
Michigan Ross specifies APA in-text format — (Tool Name, personal communication, Month Day, Year), for example (OpenAI, personal communication, September 1, 2026) — dropped directly into the sentence where the AI-assisted content appears.
LBS asks for a citation or footnote wherever AI shaped the essay, without mandating a specific format.
Formats change between cycles, and the schools that don’t require disclosure still reserve the right to screen. A full breakdown of every school’s stance is in our rundown of 2026-27 policies. Check the current, verbatim language for your target schools on our AI Policy Tracker before you write a word of your statement.
What makes a disclosure statement credible?
Four things, in order. Name the specific tool and version, “ChatGPT” is vaguer than “ChatGPT (GPT-4o),” and specificity signals you’re being precise rather than defensive. Name the specific task, not “brainstorming,” but something concrete, like testing alternate opening lines or checking whether a transition landed. State explicitly what stayed yours. And state explicitly what the AI did not do.
That last sentence is the one nobody writes, and it’s the most trust-building line in the genre. “I did not use AI to generate ideas, structure, or any sentence that appears in this essay” tells the reader more than three paragraphs of hedging about how you value authenticity. It’s a fact they can weigh, not a mood they have to take your word for.
What does a good disclosure look like?
Light use:
“I used ChatGPT (GPT-4o) to check grammar and suggest alternative phrasings for individual sentences I had already written. No AI-generated sentence appears in full; all ideas, structure, and final wording are my own.”
Moderate use:
“I used ChatGPT to pressure-test essay angles and to propose tighter versions of sentences I drafted, adopting occasional word choices but no complete sentences. The experiences, reflections, and final text are entirely my own.”
And here’s one drafted to HBS’s exact 75-word constraint, task by task:
“I used ChatGPT (GPT-4o) for three narrow tasks: generating five alternative opening lines to test against my own, checking whether my third paragraph’s transition was clear, and catching two repeated word choices across the whole draft. I kept my own opening line, every example, and my closing reflection on what the failure taught me unchanged. I did not use AI to generate ideas, structure, examples, or full sentences in this essay. The reasoning is mine.”
All three name the tool, the task, and the boundary, just in different order. The moderate example isn’t the light example with more words, it’s a different set of tasks stated with the same precision.
What should you never write in one?
A vague gesture. “AI was used to assist with this essay” states nothing a reader can evaluate, and vague is its own tell.
Overclaimed purity your own essay contradicts. If your disclosure says light editing help only and your essay reads like it was generated whole, the mismatch is the tell, not the disclosure.
Apology. A disclosure is compliance, not contrition. “I want to be transparent that I unfortunately relied on AI” reads as guilty of something the policy explicitly permits.
“If the honest version of your disclosure is uncomfortable to write, the problem is your process, not your statement.”
— Yvette Romero, MBA.AI
And the meta-point underneath all of it. No framing rescues an essay AI actually wrote. If the writing itself is another entry in essays sounding the same, the most artfully worded disclosure in the world doesn’t fix that. Fix the essay first, run it through the AI Slop Detector if you’re not sure it still sounds like you.
Keep your chats
One habit protects you more than any wording choice. Keep your prompt history. If a school ever questions authorship, your actual chat log — the prompts, the drafts, the edits — is the proof no detector can overturn. A disclosure statement is what you tell the reader. Your process log is what you could show them if you ever had to.
FAQ
Do all MBA programs require an AI disclosure statement? No. HBS, Kellogg, Michigan Ross, and LBS require formal disclosure or citation if you used AI. Others screen for it without requiring disclosure, and a few prohibit AI-written essays outright. Check your target schools’ exact language on our AI Policy Tracker before you write anything.
What if I only used AI for grammar checking, not writing? Disclose it if the school’s format asks whether you used AI at all. HBS’s yes/no box doesn’t distinguish light use from heavy use. But your statement can say exactly that. Grammar and phrasing suggestions only, no generated sentences, no generated ideas. A specific, narrow disclosure reads as more credible than a vague one, not less.
Can a disclosure statement hurt my application? A vague or overclaiming one can. A specific one, naming exactly what you used AI for and what you didn’t, reads as evidence you’re the kind of judgment-exercising leader the school is trying to admit. The disclosure itself is a writing sample.