How-ToGuide
Flagged by an AI Detector? How to Show Your Writing Process
Being told that a detector thinks your essay or report was machine-written is unsettling, especially when you wrote every word. The good news is that the best answer is usually already on your laptop: the saved versions, notes and false starts that a real piece of writing leaves behind. Collecting them carefully and presenting them plainly does more than any argument about the score.

On this page
First, freeze the evidence
Before you reply to anyone, protect what exists. Do not open the file and start rewriting sentences, do not paste the text into a new document, and do not delete old drafts to look organised. Each of those actions can break or bury the history that shows how the piece developed. If it lives only on your laptop, copy the whole folder to a USB drive or a second cloud account so there is a safe duplicate, and follow the usual advice to back up the drafts folder somewhere other than the laptop. Then note exactly what the message says: which assignment, which tool, what score and what happens next.
Where your writing record lives
Most modern editors keep more history than people realise. Google's help page for its document editor says you can open earlier versions by clicking Last edit at the top right, see who changed what, and name up to 40 versions per document. Microsoft's support guide for its cloud storage says version history works for files stored there, including word processor files and PDFs, and that versions are kept for 30 days, though a school or work administrator may have turned versioning off. If you turned on version history in advance, as described in the guide to switch on draft version history before the next deadline, you will have a longer and cleaner record.
| Evidence | Where to find it | What it shows a reviewer |
|---|---|---|
| Cloud document version history | Last edit or File, Version history | Dated growth of the text, session by session |
| Desktop file saved to cloud storage | File, Info, Version History, or the storage site | Saved states with times |
| Operating system backups | File History on Windows, Time Machine on a Mac | Hourly or daily copies of the file |
| Notes, outlines and reading lists | Notes apps, folders, notebooks | Thinking that came before the draft |
| Earlier drafts and feedback | Email, course site, messages | Changes made in response to comments |
Build your record step by step
Export a few representative versions
Open an early, a middle and the final version and save each as a PDF with its date in the file name.
Write a one-page timeline
List the dates you worked on the piece, roughly how long each session lasted and what you did: reading, outlining, first draft, restructuring, final edit. Keep it factual and short.
Gather notes and sources
Photograph handwritten notes with their dates, export outlines, and list the sources you used with where you found them. Notes that turn into an outline, and an outline that turns into a draft, are strong evidence.
Note any tools you used
If you used a grammar checker, a translation tool or an AI assistant in a way the rules allowed, write down which feature, for which part, and why. The guide to disclosing AI help you were allowed to use has wording you can adapt.
Ask about the flag itself
Politely ask which tool produced the score, what the score was and whether it is the only concern. Reviewers who understand what a detector score actually measures treat it as a prompt for questions, not as a finding.
Offer a conversation
Say you are happy to walk through the history and to talk about the argument, the sources and the choices you made. Explaining your own reasoning is often the most persuasive part.
What reviewers look for
Reviewers are not looking for a flawless process. Genuine writing is untidy: deleted sentences, moved paragraphs, a slow start and a rush near the end. What worries a reviewer is a history with one large paste and almost nothing else, or a finished text that the writer cannot explain. MIT Sloan's teaching technology team advises instructors that detection software has high error rates and can lead to false accusations, and it cites research finding that detectors wrongly flagged a majority of essays by non-native English writers. Pointing to that kind of published guidance, politely, is reasonable. So is noting that a similarity report is a different thing entirely, as explained in the guide to how a similarity report differs from an AI score.
If the history is thin
Sometimes the record is short because you wrote in one app and pasted into another. Say so plainly, offer the notes, sources and original file you do have, and offer to write a short passage on the topic in a supervised setting. For other practical guides, browse every laptop how-to on the site.
Common questions
Should I rewrite the flagged text so it scores lower?
No. Changing the text after a flag looks like hiding something and destroys the record you need. Keep the original exactly as submitted.
Is version history proof that I wrote it?
It is strong evidence of a process, not a certificate. Combined with notes, sources and your ability to explain the work, it is usually what a fair review relies on.
What if I used a grammar checker?
Say so. Tools that rewrite sentences can make text smoother, which some detectors score as machine-like. Being open about it helps.
Can I ask for a different reviewer?
Most schools and employers have an appeal or review route. Read the published policy and follow its steps if the first conversation does not resolve things.
Sources
- Google Docs Editors Help: Find what's changed in a file: opening, naming and restoring earlier document versions.
- Microsoft Support: Restore a previous version of a file stored in OneDrive: version history for stored files, kept for 30 days.
- MIT Sloan Teaching and Learning Technologies: AI Detectors Don't Work. Here's What to Do Instead.: detector error rates and false accusations.