The 72-hour playbook

Falsely Accused of Using ChatGPT? Here's Exactly What to Do.

You opened the email and your stomach dropped. The professor wrote that your paper was flagged. Maybe Turnitin returned an "AI generated" score, maybe GPTZero, maybe the professor read it and "knew." This page is the playbook for the next seventy-two hours. It is built on named, primary-source cases of students who were cleared, and on the verified limits of every detector in use today.

The seven steps

1. Do not reply for at least one sleep cycle.

Adrenaline writes terrible replies. Even a polite "I think there has been a misunderstanding" can complicate matters if it precedes the evidence. Wait. The accusation is not a deadline by itself, and any professor or honor-code office that pushes you to respond inside an hour is not following their own procedure. Take the night.

Do not use the words "I'm sorry" in any reply, in either direction. An apology can be read as an admission.

2. Screenshot everything.

Capture the original message in full, including the sender's address and any timestamps. If the message references a specific assignment, screenshot the assignment page on your learning-management system. If the accusation was made in person, write down what was said and on what date as soon as possible, while it is fresh.

3. Lock down your evidence before it changes.

This is the single most important step. Process evidence — the record of how you wrote the paper — is what cleared William Quarterman at UC Davis when his exam essay was flagged: he showed two to three hours of time-stamped Google Docs edits. The same approach worked for Louise Stivers, also at UC Davis. The full evidence guide is at how to prove you didn't use AI; the immediate steps tonight are:

• If you used Google Docs: open the document, go to Tools → Version history → See version history. Take a full-page screenshot of the panel.
• If you used Microsoft Word with cloud save: File → Info → Version History and screenshot.
• If you used Word offline: File → Info → Properties → Advanced Properties. Note the "Total Editing Time" and screenshot.
• Do not edit the document further until your case is resolved.
• If you have handwritten notes or printed drafts, photograph them with a date visible.

4. Request specific evidence — in writing.

Send one polite, factual email asking for the specific evidence the accusation is based on. If the only evidence is a detector score, ask for the screenshot of the result. You have the right to know the case against you before you respond to it.

A sample line: "Thank you for letting me know about your concern. So that I can prepare a complete response, would you please share the specific evidence the conclusion is based on, including the detector score if one was used and the date the assignment was checked? I will reply with my evidence by [date]."

5. Write your chronology.

Sit down with a coffee and write a one-page chronology of how you wrote the paper. Dates. Times. Where you sat. What you ate. What sources you consulted in what order. The details are not for the professor — they are for you. They make your story concrete and they cross-reference your version history.

6. Run the Defense Assistant.

The Defense Assistant asks five questions and produces a defense plan personalized to your jurisdiction, the named detector, your writing method, sensitive factors (ESL, disability accommodations, visa status), and stage of the process. About ninety seconds. The plan is yours to print or save. Nothing is sent anywhere.

7. Schedule the meeting only after your evidence is assembled.

A meeting before you have your evidence is a trap, even if the professor means well. Reply to schedule for a specific date — two or three working days after the accusation is reasonable. Bring printouts of the version history, your chronology, the plan, and a printed copy of the institutional integrity policy. If you can bring an advisor, do.

What the named cases tell us

The student-defense playbook is not theoretical. In 2023, William Quarterman at UC Davis was formally accused after his professor ran his exam essay through an AI detector. He defended himself with time-stamped Google Docs edit history showing the paper was written over two to three hours. He was cleared. Louise Stivers at the same institution used the same approach: she showed Google Docs version history with timestamps demonstrating she authored the paper. The charge was dropped.

In February 2026, Orion Newby won a federal ruling against Adelphi University. The judge called the AI plagiarism finding "without merit." The plagiarism finding was removed from his record. His family had paid six-figure legal costs; the win was real but it was not free.

You may not need to sue. Most cases are resolved internally when the student walks in prepared. Preparation is your job for the next seventy-two hours.

The science says the detectors are wrong often

You are not arguing against a reliable instrument. You are arguing against a flawed one whose own makers have admitted its limits.

• Turnitin's own documentation admits a sentence-level false-positive rate of about 4 percent.
• Vanderbilt University disabled Turnitin's AI detector on August 16, 2023, estimating it would have falsely flagged 750 papers per 75,000 submitted in 2022.
• Stanford researchers (Zou et al., 2023) found that seven AI detectors falsely flagged essays by non-native English writers 61.3 percent of the time on average. On 20 percent of papers, the misclassification was unanimous across detectors.
• OpenAI shut down its own AI Classifier in July 2023, citing "low rate of accuracy."
• Independent academic studies (Weber-Wulff et al., 2023) report general false-positive rates of 5 to 20 percent.

For the full data table, see AI detector false-positive rates. For your appeal, see how to appeal an AI cheating accusation.

Sources: Vanderbilt University · Stanford HAI · Tech Learning (Quarterman case) · Rolling Stone (Stivers case) · Lawsuit tracker. Full list on the sources page.