- A detector score is a guess, not proof. Cornell tells staff not to rely on it. Pittsburgh and Vanderbilt turned Turnitin's AI detector off.
- In one study of accused students, 78% had done nothing wrong.
- Do not open or edit the file. Its history is your best proof.
- Reply in writing. Stay calm. Offer to explain your work out loud.
- Check your appeal deadline today. It is often only 5 to 10 working days.
You wrote the essay yourself. You remember the late night. You remember deleting the same paragraph twice.
Then the email lands. Your work has been flagged as AI. You have to explain yourself.
Being falsely accused of AI cheating is frightening. It is also far more common than most students think. The good news is simple: these cases are very winnable. You just have to do things in the right order — and the first hour matters most.
Why detectors flag real students
An AI detector cannot see where your words came from. It only measures how predictable your writing is. Then it turns that into a percentage.
Clear writing is predictable. That is the whole problem.
Short sentences. One idea per paragraph. Simple words. That is what every teacher asks for. It is also what these tools score as "AI".
It does not mean 87% of your essay was written by a machine. It does not mean the tool is 87% sure. It is a guess about how ordinary your sentences look. Nothing more.
How often are AI detectors wrong?
Often enough that universities have started switching them off.
Students who are neurodivergent get flagged more often too. So do students who write in short, plain sentences.
This is not a fringe complaint any more. Cornell now tells staff not to treat a detector score as proof. The University of Pittsburgh and Vanderbilt have turned Turnitin's AI detector off completely. In May 2026, a family in California went to court after their son's essay on The Crucible was flagged as 76% AI.
You are not asking anyone to take your word for it. You are pointing at what universities have already decided about their own tools.
What to do in the first hour
Do not open or edit the file
Not to check it. Not to fix a typo. Every save changes the document history — and that history is what proves you wrote it. Leave it alone.
Do not reply while you are upset
An angry message becomes part of your file. Write it if it helps. Then sleep, and send a shorter one.
Get it in writing
If someone told you in person, email them to confirm what was said. You want a written record from the start.
Ask two questions
Which tool gave the score? What was the score? You are allowed to know. Often the answer shows the score is the only evidence they have.
Find your deadline
Appeal windows are short — often 5 to 10 working days. Check your student handbook today, not next week.
How to prove you didn't use AI
You cannot argue your way out of this. You show your working instead.
Put your evidence in date order, not "best first". A timeline is convincing. A pile of files is not.
The email to send (copy and paste)
Keep your first reply short. You are not making your full case yet. You are staying calm, asking for facts, and buying time.
Your first reply
Dear [name],
Thank you for letting me know. I wrote this assignment myself, and I am happy to show how.
So that I can respond properly, could you please confirm:
1. Which tool produced the score, and what the score was
2. Whether any evidence other than that score is being relied on
3. The deadline for my written response
I can provide my full version history for the document, and I am happy to discuss the work in person at your convenience.
Kind regards,
[your name] — [student number]
That email does four useful things. It denies the accusation clearly. It asks for the evidence. It offers to meet. And it puts everything in writing.
How to write your appeal
Keep it short and factual. No anger, no apology. Four parts:
The four parts of a strong appeal
- What this is about — the assignment, the date, and what you were accused of.
- Your evidence — version history, other detector results, your notes. Numbered, in date order.
- The problem with the evidence against you — a detector score is a guess, named universities have stopped using these tools, and nothing else has been offered.
- What you want — a review, and a meeting to talk through your work.
Say it once, clearly: "I wrote this work myself and I can show how." That one sentence is worth more than three paragraphs of defending yourself.
Do not say "I only used AI a little." Do not say "everyone does it." Do not apologise for something you did not do. And never offer to rewrite the essay to make it "sound less like AI" — that destroys your best evidence and looks like hiding.
If they ask you to explain your work
This is good news, not bad. A conversation is where the score stops mattering.
Be ready to say why you chose your argument, what you cut and why, where a fact came from, and what you would change now. Nobody can fake that about work they did not do.
If being put on the spot worries you, practise first. Explaining your own topic out loud is a skill, and it gets much easier the second time.
How to protect yourself from now on
None of this is your fault. But four habits make the problem disappear, because you can always show your work.
Write where history is saved
Google Docs, or Word with AutoSave on. Never write in a notes app and paste the finished essay in. No history means nothing to show.
Never paste in finished text
Even your own writing from another file arrives with no history behind it. Type it, or paste it and then really edit it there.
Keep one folder per assignment
Notes, sources, your first outline. It takes seconds and it ends arguments before they start.
If you use AI, say so
Most schools allow it for ideas or grammar. Saying so is a non-issue. Hiding it is what turns into a case.
What happens next
Most cases end in one of three ways.
At every stage, ask for decisions in writing. If a decision goes against you, ask what the next level of appeal is. There is almost always one.
Frequently Asked Questions
Can an AI detector prove I cheated?
No. It measures how predictable your writing is and turns that into a percentage. It cannot see where your words came from. That is why several universities have stopped treating these scores as proof.
How accurate are AI detectors?
Far less accurate than the marketing suggests. In one study of accused students, 78% had done nothing wrong. Even a tool that is 99.98% accurate would wrongly accuse about 80 students a year at a 20,000-student university.
Why do AI detectors flag non-native English speakers?
Writing in a second language tends to use simpler, more regular sentences. Detectors read that pattern as machine-written. Stanford researchers measured false-positive rates of about 61% on essays by non-native English writers.
Should I rewrite my essay so it sounds less like AI?
No. Never change a document that is being reviewed. Editing destroys the version history that is your best evidence, and it looks like you are hiding something.
How do I prove I wrote my essay myself?
Version history is the strongest proof. Google Docs saves timestamped versions under File → Version history. Word records changes through Track Changes and AutoSave. An essay that grew over several days looks nothing like one pasted in at once.
What if I used AI only for grammar or ideas?
Say so straight away. Most schools allow assistive use. Being open about a permitted tool is a completely different conversation from being accused of hiding it.
What if my school ignores my evidence?
Ask in writing for the exact policy the decision relies on, and for the detector's documented accuracy. Then use the formal appeals process rather than arguing informally. Your students' union can help you write it, free of charge.
Can I be expelled for something a detector flagged?
A first allegation based only on a detector score rarely leads to the most serious outcomes, especially once evidence is provided. Deadlines matter more than panic here — find your appeal window and reply inside it.
The bottom line
An accusation is frightening. But a detector score is a guess, and a guess loses to a documented process every time.
Leave the file alone. Collect your history. Put your evidence in date order. Offer to talk about your own work.
Then change how you write next time. The best defence is a document that can show exactly how it was made.
References & Sources
Reported false-positive rates, and the study finding that 78% of reviewed accusations were mistaken — from published analyses of AI-detection accuracy and academic-integrity casework, 2026.
Stanford research on raised false-positive rates (about 61%) for essays written by non-native English speakers, widely cited in student appeals.
Institutional responses: Cornell guidance against using AI-detector scores as the basis for a violation; the University of Pittsburgh and Vanderbilt disabling Turnitin's AI detector.
Civil suit filed May 2026 after a Turnitin AI flag of 76% on a high-school essay in Palo Alto, California.
Guidance current as of September 5, 2026. This article is general information, not legal advice. Policies and appeal deadlines vary — always check your own institution's academic-integrity rules, which often allow only 5 to 10 working days to respond.