Falsely accused of using AI? What to do, step by step

In the first hour

Start by slowing down. A flag opens a question, and what you do in the first hour can make that question easier or harder to answer.

• Read the message twice and write down any deadline for replying.
• Leave the flagged document exactly as it is. Its history is your best evidence, and edits made now blur it.
• Keep the text away from rewriting tools, and do not count on a second detector to settle things. Another score is another estimate.
• Do not admit to something you did not do to make the problem go away, and do not reply in anger. A short, polite request for the details is enough for now.

Then find the rules. Your syllabus and your institution’s academic integrity procedure say who decides and what the steps are. They may also say whether you can bring someone with you to a meeting.

What a detector score is, and is not

An AI detector reads a text and estimates how closely its wording matches patterns the tool links with AI writing. A detector that scores a finished text has no record of what you typed and no view of how you worked. Its output, a percentage or a label, is a prediction about the text. It is not evidence of who wrote it.

What the number counts depends on the tool. Turnitin, for example, describes its figure as the share of a document that AI tools may have generated or modified. Ask which tool was used and what its number means.

The companies that build detectors say similar things themselves.

OpenAI released its own AI text classifier on January 31, 2023. At launch it said the tool correctly identified 26% of AI-written text and wrongly labeled human-written text as AI-written 9% of the time. It also said the tool should not be used as a primary decision-making tool. A note on that announcement now says that as of July 20, 2023 the classifier is no longer available, “due to its low rate of accuracy”.

Turnitin builds an AI writing detector for educators. Its own FAQ, as we read it on November 13, 2026, says Turnitin does not make a determination of misconduct, and that the percentage “should not be used as the sole basis for action” by instructors. The same page says Turnitin works to keep its false positive rate, meaning fully human writing identified as AI, under 1% for documents with over 20% of AI writing. A low rate is still not zero.

The errors do not fall evenly. In a study published in the journal Patterns in July 2023, Liang and colleagues ran seven detectors on 91 essays from the TOEFL test of English as a foreign language, written by non-native speakers, and on 88 essays by US eighth-graders. The detectors classified the eighth-graders’ essays accurately and wrongly labeled more than half of the TOEFL essays as AI-generated, an average false-positive rate of 61.3%. That study is small and dates from 2023, so it does not tell you how often today’s tools are wrong. It does show how plain, careful writing in a second language can be flagged.

None of this proves that a particular flag is wrong. It means a score cannot carry a case alone, and the rest of this guide is about the evidence that can help.

1. Gather the record of how you wrote it

The best answer to “did you write this?” is a record of the writing as it happened. Collect what you have before any meeting, and copy it somewhere safe without changing the originals.

Version history is often the most useful item. Google Docs keeps a history of earlier versions, with who made each change, for anyone who can edit the file. Word keeps a version history too, but only for files stored in OneDrive or SharePoint, so a document saved only on your own computer will not have one. We checked both help pages on November 13, 2026.

If you have no version history, say so plainly and rely on the other items. A missing history proves nothing either way.

If you used an AI tool for something your course allows, such as a grammar pass, Edit Check compares your original with the AI’s version in your browser and lists what changed. That helps you show how much of the text was yours.

2. Ask for the evidence and the process in writing

You cannot answer a concern you have not seen. Ask, politely and in writing, for the evidence and for the rule that applies. Ask too what the steps of the review are and when you need to reply. Writing gives both sides a record, and it gives you time to think.

Here is an email to adapt. Fill in the blanks in double braces and delete any line that does not fit your case. For the evidence blank, list what you gathered in step 1. For the AI blank, write either “I did not use any AI tool for this work” or, if your course allows AI, exactly what you used it for.

Subject: {{assignment_name}} in {{course_name}}: request for the evidence and the review steps

Dear {{instructor_name}},

Thank you for telling me about your concern with my {{assignment_name}}, which I handed in on {{date_submitted}}. I wrote this work myself, and I would like to help clear this up as soon as possible.

So that I can prepare, could you please send me:
1. The report or other evidence the concern is based on, including the name of the tool that was used and the result it gave.
2. The part of the course or institution policy that applies.
3. The steps of the review and who makes the decision.
4. The date by which I should reply.
5. Whether I may bring someone with me to a meeting.

I can show how I wrote the work. I have {{evidence_you_have}}, and I am glad to go through it with you and to answer questions about the content in person.

{{one_sentence_about_ai_use}}

I am free on {{times_you_are_free}}. Thank you for your time.

Kind regards,
{{your_name}}
{{student_number}}

Send it from your school account and keep a copy. Write every message as if it will be read aloud later in the process.

3. Explain your work aloud

In a meeting, the most useful thing you can do is talk about the work the way its author can. Bring the record from step 1 and be ready to:

• Tell the story of the work, from how you chose the topic to what you changed along the way.
• Explain any paragraph in your own words and define the terms you used.
• Say where each main claim comes from and what the source says.
• Open the version history and scroll through the dates.

Answer what you are asked, briefly and honestly. If you do not remember a detail, say so. If the policy lets you bring a support person, such as an adviser or a student representative, consider it.

Rehearsing helps. The best partner is a friend who asks hard questions. If nobody is free and your course does not ban AI tools, a chatbot can play the instructor with the prompt below. Use a new chat, and do not let it change your text.

Help me rehearse a meeting with my instructor. My {{assignment_type}} for {{course}} was flagged by an AI detector. I wrote it myself, and my aim is to explain it clearly and calmly in my own words.

Play the instructor. Use only my text and my notes below.

<my_text>
{{your_text}}
</my_text>

<my_notes_and_sources>
{{your_notes_and_sources}}
</my_notes_and_sources>

Ask me 8 questions, one at a time, and wait for my answer after each one. Mix these kinds:
- Why I made a choice, for example "Why did you open with this argument?"
- What a term or sentence in my text means.
- Where a claim comes from and what the source says.
- How I planned and drafted the work.

After each answer, tell me in one or two sentences whether it was clear and specific, and what a listener might still want to know. Then ask the next question.

Do not rewrite my text or write answers for me. Do not give an opinion on whether the text looks AI-written. A good session leaves me able to talk through every paragraph without reading from it. If a question depends on something that is not in my text or notes, ask me about it instead of assuming.

Format: plain text, short messages, no headings.

If AI was part of your process

Be exact about it. If you used AI in a way your course allows, say what it did and what you did. Show the chat if you still have it. A precise account is easier to check than a flat “I never use AI”. The AI Use Statement tool writes a short statement of how you used AI that you can give to a teacher.

If you used AI in a way the rule did not allow, the steps in this guide will not change that. Saying so early is the honest course, and what follows is for your institution to decide.

If this happens at work

An employee whose report is suspected of being AI-written is in a similar position under different rules. Ask what the concern is based on and which company policy applies. Show your drafts and the sources you worked from. If the policy allowed the tool you used, say how you used it. Your employer decides under its own policy, and a union representative or an employment adviser can tell you where you stand.

What to do differently next time

• Write in an app that keeps version history, from the first line. Large blocks pasted in from elsewhere arrive in the history all at once, with no trail behind them.
• Keep your outline and earlier drafts until the grade is final.
• Read each course’s AI rule at the start, and ask in writing when it is unclear.
• If the course allows AI, say how you used it when you hand the work in, not after a question comes up.

Some writing habits make readers think of AI, such as stock phrases and a neat summary at the end of every piece. The Robot-Speak Checker shows those habits in your own draft. It says plainly that it is not a detector, and it cannot make a text pass one. It helps you sound like yourself, and the guide to AI words to avoid explains the habits behind it.

Stay away from tools that promise to beat detectors. They change your words and leave you with a text that is less your own. If you use AI to study, our guide on how to use AI to study without cheating covers which uses to ask about first.

Related

• How to use AI to study without cheatingSix habits that keep the thinking yours, with a tutor prompt to copy.Guide
• Robot-Speak CheckerSee the writing habits readers link with AI. It is not a detector.Free tool
• Edit CheckSee exactly what an AI changed in your own text.Free tool
• AI words to avoidThe words and habits that make writing sound like AI, and what to write instead.Guide