Can You Use ChatGPT on a Supplementary Application?

What the programs say, in their own words

Ontario has no single AI policy for applications. Each program writes its own, and some have not written one at all. This is what is public as of September 2026, for Fall 2027 entry. Confirm each on the linked page before you rely on it.

• McMaster Health Sci (BHSc) is the strictest. Its supplementary application FAQ says that posting your answers online, use of paid admissions consultants, plagiarism, “use of generative artificial intelligence, or otherwise working with others to prepare your responses is not permitted and may result in removal from the applicant pool or withdrawal of an offer of admission.” It goes further than most: “Your responses to the questions must be entirely your own. Discussion with others, including current program students, is not permitted.” Read that as no AI, no consultant, no parent editing, no friend proofreading. The same FAQ says a typo or two will not hurt your score, which tells you what they would rather see than a polished paragraph someone else touched.
• McMaster Engineering’s supplementary application page requires that all answers on Kira “are entirely your own original work,” and says: “Do not use AI tools, other people, or any external sources to write or generate your responses.”
• Waterloo’s AIF page says “You must complete and submit the AIF on your own,” and: “You’re responsible for the answers you provide. If you use AI to create answers that are untrue, incorrect, or plagiarized, we reserve the right to deny admission or withdraw your Offer of Admission.” Notice the wording. Waterloo does not say “never touch AI”; it says you own whatever comes out, and false or plagiarized content is grounds for losing the offer. The same page says the AIF is “reviewed both by people who have been trained on reviewing supplementary information forms and by artificial intelligence (AI),” with a second human scorer when the two disagree significantly. If you have heard that Waterloo is letting AI grade the AIFs, that is where it comes from. It is true, it is published, and it is a scoring aid with a human check, not an AI detector deciding your fate.
• OUAC’s undergraduate declaration, which every Ontario applicant agrees to, includes that “any applicable autobiographic and personal submissions are true, accurate and authored solely and entirely by me.” It does not name AI. It does not need to. A ghostwritten supp app is not “authored solely and entirely” by you, whoever or whatever the ghost was.
• Ivey AEO, Rotman Commerce, Queen’s and U of T Engineering publish no AI-specific statement on their applicant pages that we could find. That is not permission. Rotman says its written and video responses are its chance to get to know you; the OUAC declaration applies to all of them; and Ivey’s essays go to a reviewer who reads hundreds of them and will recognise the ChatGPT cadence in a sentence.

Why your own writing gets flagged

Now the other fear, and it is the more common one. You wrote the thing yourself, you pasted it into a free detector out of paranoia, and it said 50 or 66 percent AI. First, take a breath: none of the programs above says it runs your supp app through a commercial AI detector, and the people who actually score it are reading for content. Second, here is why the detector is wrong about you.

Detectors mostly measure two things. Perplexity: how predictable each next word is. Burstiness: how much your sentence length and rhythm vary. Language models write predictable words in even sentences, so text with low perplexity and low burstiness scores as “AI.” The problem is that a nervous 17-year-old writing a formal 1,500-character answer for a program they desperately want also produces predictable words in even sentences. You reached for the safe phrase (“this experience taught me the importance of”), you kept every sentence to a tidy medium length because a teacher once told you to, and you removed every contraction and every odd detail because it felt unprofessional. You wrote the way a model writes, on purpose, because school trained you to. The detector cannot tell the difference between a model and a student imitating one, and it is not built to. OpenAI shut down its own AI-text classifier in 2023 for low accuracy; the free ones you are pasting into are not better.

So the fix for “my own essay looks like AI” and the fix for “my essay is generic” are the same fix, which is convenient. A reader who scores you highly and a detector that clears you are both reacting to the same thing: specificity and a human rhythm.

Should you run your supp app through an AI checker?

No, or at least not for the reason you want to. The free detectors score how predictable your prose is, and a careful 17-year-old writing formal sentences for a university scores as predictable. That is the 66 percent. None of the programs above says it runs applications through a commercial detector; the ones that read for AI are reading for the cadence, the generic claims and the missing specifics, which a detector does not measure and a human does. If you want a check, use the checklist below, which tests for the things reviewers actually notice, or ask a reader to flag any sentence that does not sound like you. AcceptMe’s essay coach does that as part of its notes: it marks adult voice, claims without evidence and padding, and it does not produce a rewrite, so there is nothing to paste back in.

The “sound like a person” checklist

Run every paragraph through these six. Each one comes with a before and after, because the rule alone is not enough.

1. Replace every abstract noun with the thing it stands for.
   Before

   This experience taught me the importance of effective communication and adaptability in a team setting.

   After

   I learned to text the group a two-line plan before every shift instead of hoping people would ask.

2. Put a number where an adjective is.
   Before

   I helped significantly increase attendance at our club’s events over the year.

   After

   Attendance went from about 12 to 40 by March, mostly because we moved meetings to lunch.

3. Vary the length. One short sentence per paragraph, minimum.
   Before

   The project was challenging because of the timeline. The team was not always aligned on the goals. I took steps to improve the situation over time.

   After

   We had nine days and no plan. Two of us wanted a video, three wanted a poster, and nobody had said so out loud until I asked the question in the group chat that Tuesday night.

4. Name a person, a place or a date. Models avoid specifics because they do not have any. You do.
   Before

   A mentor gave me valuable feedback that changed my approach.

   After

   Mr. Osei handed the lab back and said, “You answered a different question than the one I asked.” He was right.

5. Admit one thing that did not work. A model writes a clean arc; a person remembers the part that went wrong.
   Before

   Through perseverance and dedication, the fundraiser was ultimately a success.

   After

   The first bake sale made $46 and I had to throw out two trays. The second one, with a pre-order sheet, made $310.

6. End on what changed in you, stated plainly, not on a moral.
   Before

   Ultimately, this journey shaped me into a more resilient and empathetic individual, ready to face future challenges.

   After

   I still hate asking for help. I now do it on day two instead of week three.

Read your draft aloud once. Anywhere you would be embarrassed to say the sentence to a friend is a sentence a model could have written. Rewrite those and only those.

A sane position on using AI at all

There are three different things people mean by “using ChatGPT,” and lumping them together is why the question never gets a straight answer.

• Brainstorming. “Ask me ten questions about my part-time job that might surface a story.” Nothing it produces goes on the page. This is the same as talking to a friend, and only McMaster Health Sci bars even that (it bars discussion with anyone). Everywhere else, it is fine and honestly a bit useful.
• Editing your draft. Pasting your paragraph in and asking “what is unclear?” is feedback. Pasting it in and saying “make this better” and accepting the rewrite is ghostwriting, because the words that come back are not yours, and they will be the same words that come back for two thousand other applicants. If you do this, you will sound like everyone else who did this, which is the one outcome a supp app exists to prevent.
• Generating. “Write my Ivey essay about being captain.” This is what every policy above is aimed at. It is also, separately, a bad strategy: the output has no numbers, no names, no failure, and the flat rhythm that both a reviewer and a detector notice in seconds. A mom who rewrites your U of T supp app in ChatGPT has done exactly this, with love, and it needs to be undone before you submit.

Feedback on your own writing is a different category from writing done for you, and it is the category AcceptMe’s essay coach lives in. You paste in the answer you wrote; it returns a verdict paragraph and notes tied to your exact sentences: this line works, this one is a claim without evidence, this one repeats the question back. It does not produce replacement text. That is not because generating it would be hard; it is because the moment a tool writes the sentence, it stops being your supp app. Write-it-for-me tools have a place in the world. An application whose entire purpose is to let a reader hear you is not that place.

To be fair to the chatbot, it is not dumber than our coach. What it lacks is context. It does not know the works, tighten, fix approach, it does not have the question, the limit and the program’s own wording in front of it, and it does not know what a reviewer is actually scoring, so what you get back depends on how well you happened to prompt it that evening. A tool built for the job carries that context for you, which is the whole point: you get each piece right without having to become an expert in supp apps first.

If you are staring at a 66 percent right now

1. Close the detector. It is not the reader, and no Ontario program says it uses one.
2. Run the six-item checklist above on the two most generic paragraphs. That is usually the opening and the last one.
3. Keep your drafts. Google Docs version history, dated, is the best evidence that exists that a human wrote something over eleven evenings. If a program ever asks, you have it.
4. For McMaster Health Sci specifically, do not ask anyone to proofread. For everything else, one person reading for clarity is fine; nobody rewriting is the line.

Where to go from here

If the reason you were tempted by ChatGPT is that you cannot think of anything to write about, the no-ECs guide shows how a part-time job, a sibling and a hobby each turn into a scored competency story. If you are new to all of this, the what is a supp app guide explains who needs one and when each is due. If it is the two Ivey essays specifically, the AEO essays guide breaks down both prompts and how to keep them from repeating your video. If you were about to pay a person instead, the comparison of supplementary application review services covers what each one reads and what the programs allow. Then get your own draft marked up: AcceptMe’s essay coach reads the answer you wrote and returns line-by-line notes on what works and what to fix, without writing a word of it for you.