Falsely Accused of Using AI? A Calm, Practical Guide for Students

What should you do first if you're accused of using AI?

Slow down before you reply. An email saying your work was flagged by an AI detector feels like a verdict, but in most institutions it is the start of a process, not the end of one. Your first job is to protect your evidence and understand what you are actually being asked.

Do not delete, edit or re-upload the assignment. Do not rewrite the file to make it look different, and do not run it through a paraphrasing or "humanizer" tool. Changing the document now can make your version history harder to read and can look like you are hiding something, even when you are not.

Then read the message closely. Note who sent it, what it claims, what it asks you to do and by when. Some institutions start with an informal conversation with the instructor; others go straight to a formal referral. The next steps differ, so work out which one you are in before you respond.

How do you prove you wrote it yourself?

You usually cannot prove a negative, but you can show your process. The strongest evidence that you wrote something is a record of you writing it: drafts that change over time, notes, sources you read, and the edits you made along the way. Gather it before any meeting.

Can an AI detector score prove you used AI?

No. AI detectors estimate how statistically similar a text is to machine-written text. They do not observe how the text was written, and every one of them makes mistakes. A score is a signal that something may be worth a closer look, not proof that misconduct happened.

The vendors say this themselves. Turnitin states that because its false positive rate is not zero, instructors need to apply professional judgment and that it does not make a determination of misconduct. In a 2023 update, Turnitin reported a document-level false positive rate below 1% for documents with more than 20% AI writing, a sentence-level rate of about 4%, and a higher incidence of false positives when less than 20% of a document is flagged.

Independent research is less reassuring. A Stanford-led study by Liang and colleagues, published in Patterns in 2023, found that seven AI detectors misclassified 61% of TOEFL essays by non-native English writers as AI-generated on average, while essays by US eighth-graders were rarely flagged. A separate review of 14 tools by Weber-Wulff and colleagues concluded that the available detection tools were neither accurate nor reliable.

For a fuller explanation of why human writing gets flagged, and what published false-positive rates actually measure, see our guide to AI detector false positives.

What do universities say about AI detection?

Several universities have publicly limited or switched off AI detection, which is useful context when you talk to your instructor. You are not asking them to ignore a concern; you are pointing out that their own sector treats these scores with caution.

Vanderbilt University disabled Turnitin's AI detector in August 2023. It calculated that even a 1% false positive rate, applied to the roughly 75,000 papers it submitted in a year, could have wrongly flagged around 750 student papers, and it cited research showing detectors are more likely to flag non-native English writing.

Curtin University in Australia disabled Turnitin's AI writing detection from 1 January 2026, while keeping its regular text-matching checks. MIT Sloan's teaching technology team advises instructors that AI detection software has high error rates and can lead to false accusations, and recommends clear policies and conversations with students instead.

None of this means your institution must drop its case. It does mean a reasonable decision-maker should look at more than a percentage, and you are entitled to ask that they do.

What to do if you're accused of using AI: step by step

Procedures differ between schools and countries, so treat this as a checklist to adapt, not a script. Your institution's written policy always takes priority.

1. 1

   Keep everything as it is. Do not delete, edit or re-upload the flagged file. Save a copy of the accusation email and any report or screenshot you were shown.

2. 2

   Export your version history. Open the document in Google Docs or Word on OneDrive and record the version history, naming key drafts. Gather notes, outlines, sources and earlier copies.

3. 3

   Find your institution's academic integrity procedure. Read the policy that applies to your course. Note the stages, deadlines, what counts as evidence, and whether you can bring a support person or adviser.

4. 4

   Ask which tool and threshold were used. Request the detector name, the score, the highlighted passages, and whether any other evidence was considered. Ask what score or rule triggered the referral.

5. 5

   Request a human review. Ask in writing that the decision be based on a review of your work and process by a person, not on the detector score alone. Offer to discuss your work in a meeting.

6. 6

   Prepare to explain your work. Reread your essay and sources. Be ready to explain your argument, how you found your sources and why you wrote particular passages the way you did.

7. 7

   Get support from your student advice service. Contact your students' union, ombuds office, academic adviser or student advocacy service. They know the local process and can attend meetings in many institutions.

8. 8

   Keep a written record. After each meeting or call, send a short email summarizing what was said and agreed. Keep copies of every message and decision.

If the outcome goes against you, check whether your institution has an appeal stage and its deadline. Appeals usually need to show a procedural error or new evidence, so your notes and version history matter here too.

What questions should you ask about the AI detector?

You are entitled to understand the evidence against you. Ask these questions politely and in writing, and keep the answers.

Are some students more likely to be falsely accused?

Yes, the research points that way. Detectors look for predictable word choice and sentence structure. Writers who use a smaller, careful vocabulary, which is common when English is a second language, can look "predictable" to a model. That is the pattern the Liang study found, and it is the reason Vanderbilt gave for caution.

Highly formulaic assignments carry similar risk: lab reports, literature-review summaries, short reflective pieces that follow a template, and very short answers. Turnitin raised its minimum length for AI scoring from 150 to 300 words in 2023 because shorter texts were less reliable, and it reported more false positives in opening and closing sentences.

If any of this describes your writing, say so calmly in your meeting. It is a known limitation of the technology, not an excuse.

How can you protect yourself on future assignments?

The best protection is a visible writing process. Draft in Google Docs or in Word saved to OneDrive so version history is kept automatically. Keep your notes and sources in one folder per assignment. If your course permits AI tools for some tasks, keep a short note of what you used and for what, and check the course policy before you rely on it.

Some students run a draft through an AI detector before submitting, simply to see which sentences a tool might highlight and to be ready to discuss them. That is reasonable if you treat the result as a prompt to reread, not a verdict. Plagino's AI detector shows sentence-level highlights in a PDF report, and our benchmark publishes its measured false-positive rate alongside other tools, including ones that beat it on detection. A clean result from any tool does not guarantee your school's tool will agree.

This guide is general information, not legal advice. If your case could affect your visa, funding or degree, contact your institution's student advice service early, and seek independent advice where the stakes are high.

Frequently asked questions