Accused of Using AI When You Did Not: What to Do

The email arrives with a subject line like Academic Integrity Concern or Meeting Request: Regarding Your Submission, and the first line inside it tells you that your instructor believes you used artificial intelligence on work you wrote yourself. Getting accused of using AI on essay work you actually wrote is disorienting in a specific way: there is no single fact you can hold up that disproves a percentage, and the instinct to panic, argue, or over-explain in the first message you send is exactly the wrong move. The fear is real and reasonable. The response to it still has to be procedural rather than emotional, because procedure is what the next few weeks will actually run on.

This is not about whether the detector was right. Plenty of writing gets flagged for reasons that have nothing to do with authorship, and TextPulse has covered how accurate these tools actually are in detail elsewhere. What matters in the next two days is procedure: what you say, what you ask for, and what you find out before committing to a version of events you might need to walk back later. Get the order right and you protect yourself regardless of how the underlying question eventually gets resolved.

Accused of Using AI on Essay Work? Control the First Reaction

Wait before you respond, and keep any early message short. Anger reads clearly in writing, and an academic integrity office reads a great deal of it. A reply sent within the first hour, especially one calling the process unfair or the professor incompetent, becomes part of your own file whether or not you are right to feel that way.

Take the time you are actually given. Most instructors and integrity offices expect a reply within a few days, not a few minutes, and a short holding message, noting that you received the notice and are preparing a response, buys you room to think without looking evasive. Read the message twice before drafting anything, and note down exactly what it claims: which assignment, which tool it names, and what evidence, if any, it already cites. If it names a deadline for your response, write that date down somewhere separate from the email itself, since missing it can turn a disputed accusation into an uncontested one by default.

Do Not Admit to Anything You Did Not Do

Say only what actually happened, and resist the urge to sound cooperative by admitting to more than that. A surprising number of false-accusation cases go wrong at exactly this step. A student, rattled by an unfamiliar process, says something like I did use Grammarly to check it or I might have looked something up on ChatGPT for an idea, meaning something far short of writing the essay with it.

That sentence, once written down, becomes the one fact in the file that is not in dispute, and everything else gets read through it. Answer only what is actually asked, and answer it precisely: if you used a grammar checker, a citation tool, or a spell checker, say exactly that and nothing more, because vague cooperativeness reads as evasion just as easily as silence does. If you are not sure whether a tool you used counts as the kind of AI use the policy actually means, say what you used and ask directly, rather than guessing and committing to an answer you might need to correct later.

What Evidence Should You Request in Writing?

A flagged percentage on its own is not an accusation you can actually respond to, because it does not say what triggered it. Ask, in writing, for the full report rather than the summary score: which sentences or sections were flagged, what threshold your institution uses before a score counts as a concern, and exactly which system produced the number.

Turnitin's own materials are explicit that the AI writing report and the similarity, or source matching, report are separate systems measuring separate things, and a document can score high on one while scoring low on the other. If the message you received does not say which report it is citing, that is the first clarifying question to ask, since flagged for AI and flagged for matching a source are different accusations that call for different responses. Ask specifically for a copy of the actual report, not just a description of it. Turnitin does not show the AI report to students by default, so an instructor citing a score is usually describing something you have never seen yourself, and someone with access has to export it and send it to you directly.

Put the request in an email rather than raising it only in conversation, so a record exists that you asked. A reasonable process should be able to produce this. Reluctance to share the actual evidence behind a percentage is itself worth noting, particularly if the matter later reaches a formal hearing where you will want to show exactly what you requested and when.

Keep a copy of everything you receive and everything you send from this point on, in a folder you control rather than only in your university email account. If the case moves to a formal hearing weeks or months later, remembering exactly what was said and when matters more than it will feel like it does today.

Which Policy and Which Procedure Applies to You?

Every institution runs this differently, and the label attached to your case matters more than it sounds like it should. Ask directly: is this an informal conversation between you and your instructor, or has it already been referred to an academic integrity office as a formal case? The two follow different rules, carry different stakes, and give you different rights to the evidence against you. Do not assume the answer from tone alone. A friendly-sounding email can still be the opening step of a formal process, and a stern one can still turn out to be an instructor working through a personal concern before anything is filed anywhere.

	Informal conversation	Formal referral
Who is involved	Your instructor, no case file yet	An integrity office or committee, a case number
What is typically at stake	A grade discussion, sometimes nothing further	A transcript notation, suspension, or worse in serious cases
Your right to the evidence	Ask anyway, even if it is not formally required	Usually a documented right to see the full report
What to do first	Ask which category this is, in writing	Request the written procedure by name and read it before you reply

Once you know which one you are in, find the actual policy document itself, not a summary of it that someone gives you verbally. Most universities publish this openly, and TextPulse's overview of how AI policy actually works across institutions is a reasonable place to start if your own university's version is hard to track down. The policy usually states what counts as evidence, who decides, and what your right to respond looks like on paper, and each of those details changes what your next message should say.

Can a Detector Score Be the Sole Evidence Against You?

Usually not, at least on paper. Turnitin's own guidance to institutions states plainly that its AI writing model may not always be accurate and should not be used as the sole basis for adverse actions against a student. Michigan State's own guidance repeats the same principle in almost the same words. False positives are a documented pattern across independently tested detectors, and if your case rests on a score with nothing else behind it, you are entitled to say so.

This is not a technicality you are inventing to defend yourself. Vanderbilt disabled Turnitin's AI detector entirely in 2023, citing the vendor's own published false positive rate and estimating that, across the roughly 75,000 papers it submits annually, hundreds of student papers could have been mislabeled. The University of Texas at Austin, Northwestern, and Montclair State made similar decisions around the same period. None of that proves your specific case is wrong. It shows that the number by itself has never been treated as sufficient, even by universities that pay for these tools.

This matters even more if English is not your first language. Independent testing has repeatedly found detectors misreading fluent second-language academic writing at far higher rates than native writing, and TextPulse's page for ESL academic writers goes into exactly why that happens. If that applies to you, say so in your written response, and ask specifically whether that documented pattern was accounted for before your score was treated as meaningful.

What Happens After the First 48 Hours

Once you have the evidence, the policy, and a clear sense of which process you are in, the next stage is usually a written response, sometimes a meeting, sometimes both. What you say in that meeting matters less than what you can show: a timeline, a process, a paper trail that existed before anyone accused you of anything.

The two practical next steps have pages of their own: assembling evidence that you did not use AI, and writing an appeal letter that answers the score rather than the accusation.

That paper trail is worth building whether or not you ever need it again. The habits that make a false accusation survivable, saving drafts as you go, keeping your search history, being able to walk someone through your own reasoning line by line, cost nothing to start today and a great deal to reconstruct after the fact. Start now, quietly, regardless of how this particular case turns out. What that evidence actually looks like, and how much weight each kind of it can carry, is worth understanding in more detail before you ever need it under pressure.

Related research: the findings above are examined at scale in Non-Native English Writing and the False Positives of Stylometric AI Text Classification, a TextPulse Research working paper with open data, code and a citable DOI.