Accused of using AI when you didn’t

Find out what the accusation rests on, and ask in writing so that you have the answer in writing:

• –Which tool was used, what it said, and about which passages.
• –What the rule is that you are said to have broken, and where it is written.
• –What would settle it: what evidence they will look at, and who decides.

If it is a detector score, it is worth knowing that the people who make and buy these tools do not claim they are always right. When Vanderbilt University turned off Turnitin’s AI detector in 2023, it pointed out that a 1% false positive rate, the figure Turnitin gave at launch, would have meant around 750 of the 75,000 papers its students submitted the year before being wrongly labelled (Vanderbilt’s guidance). See why detectors flag human writing.

Anything made while you wrote it, with its date on it:

• –The document’s version history, if you wrote in Google Docs or in Word on OneDrive. Share the original, not a copy, which starts with no history.
• –Your notes, outline, earlier drafts and sources, and when you made or found each.
• –Your browser history for the days you worked on it, if it shows the research.
• –A plain account of when and where you wrote it, over how many sittings.

If you wrote it in Google Docs, Draftback, a Chrome extension, replays a document’s revision history like a film, and works on documents written before it was installed (draftback.com). Many teachers already use it. More on this in the guide to Google Docs version history.

Briefly, calmly and in writing. Say that you wrote it, where and when, and attach or point to the evidence. Offer to talk the piece through: why it is built the way it is, what you cut, what you would change. If it would help, offer to write something comparable where they can see you do it.

Do not rewrite the piece to make a detector happier, and do not run it through a tool that promises to make text read as human. It changes the score, not what happened, and a piece that changed after the accusation is harder to defend than one that did not.

Most schools and universities have a written academic integrity process, and it usually says what evidence is needed and how you can respond to it. Read it, ask for a meeting if one is not offered, and ask whether there is an adviser or ombudsperson who can come with you. Bring your evidence in order, with dates. If the score came from Turnitin, see what Turnitin itself says about it.

Showing your teacher you wrote it covers the next assignment.

An editor who turns a piece down as “40% AI” is usually reading a score from a tool, and can be sent the same evidence a teacher would be. A reader who says so in public is harder: a statement that you wrote it is only a claim, and an argument about a score rarely ends. On Substack, where readers can now run a scan themselves, see flagged as AI on Substack.

You cannot stop somebody running a detector. You can have a record ready when they do. iwrote records how a piece is written as you write it in the browser, and certifies it with a short link that shows a reader how it was written and lets them check that record themselves. It covers writing done on iwrote from the first word, so it cannot vouch for a piece already handed in or published.

Start the next piece on iwrote, or read about the other ways to keep a record.