Accused of Using AI? How to Prove You Wrote It

Loch Ness · HUMAN WRITING · Proving authorship

Part of the guide: Do AI Detectors Actually Work? What the Research Says

By Camille G., founder of Loch Ness · Published September 29, 2026 · Last updated September 29, 2026

Short answer: don’t argue with the score, show your process. The version history of your document, your drafts and notes, and your ability to talk through your own choices are far stronger evidence than any detector percentage. Here’s the order I’d go in, from the quickest win to the strongest proof.

Human Origin Label · For writers The best defense is a record someone else has checked. Walk a reviewer through your drafts once, live or in a screen recording. You get a public verification page you can point to next time. Get verified, it’s free → Not ready for a review? Get the free self-declared badge →

Step 1: ask what the accusation is based on

Calmly ask whether there’s anything beyond a detector score. Often there isn’t. It’s fair to point out that independent research has found these tools neither accurate nor reliable (Weber-Wulff et al., 2023), and that they wrongly flag a large share of human writing by non-native speakers (Liang et al., 2023). Keep it factual. You’re not attacking anyone, you’re asking for the evidence to be weighed properly.

Step 2: open your version history

This is usually the strongest single piece of evidence you have. Google Docs keeps a history of earlier versions and shows who changed what and when; anyone with edit access can browse it (Google Docs Help). In Word, version history works when the file is saved in OneDrive or SharePoint (Microsoft Support).

A real writing history looks messy: sentences typed, deleted, moved, rewritten over days. A text pasted in one block at 11:52 p.m. looks very different. How to read it, share it and present it well: using version history as proof.

Step 3: gather the rest of your paper trail

Earlier drafts, handwritten notes, outlines, research tabs, the sources you read, messages where you discussed the piece. Each one alone is small. Together they show a process that only a person working on the text over time would leave behind. What to keep, even for work you’ve already finished: drafts, notes and outlines as a paper trail.

Step 4: offer to talk it through

Offer to explain your text out loud: why you chose this example, what you cut, what you’d change. People who wrote something can usually discuss it in depth without preparation. It’s simple, and in practice it’s one of the most convincing things you can do.

If it’s a client, not a teacher

Freelance writers face the same accusation with money on the line. The steps are the same, but the tone and the timing are different, and prevention matters even more: how to reassure a client about AI.

Your evidence at a glance

Evidence	Strength	Where to find it
Version history	Strong	Google Docs, or Word with OneDrive/SharePoint
Drafts and notes	Medium to strong, together	Your files, notebooks, email
Talking through the text	Strong in the moment	A short conversation
An independent review, done before any accusation	Strong, and public	A reviewed label with a verification page
Another detector score saying “human”	Weak	Same tools, same flaws

Make it easier next time

Write in a tool that keeps history, keep your notes, and if your work is public, consider having your process checked once by someone independent. That’s what I built Human Origin for: a review of how you write, and a public page anyone can check before a question even comes up.

Get your writing process reviewed once, and link to it for good. Apply for a free review →

Why the accusation happened in the first place: why AI detectors flag human writing. And if you did use AI for part of your work, here’s where the line is between help and ghostwriting.

Full disclosure: Human Origin, the label mentioned on this page, is something I created. I’m flagging it so you can weigh what I say about it. The facts above rest on the sources below.

Sources

• Weber-Wulff et al. — Testing of detection tools for AI-generated text (2023), via ERIC
• Liang et al. — GPT detectors are biased against non-native English writers, Patterns (2023)
• Google Docs Editors Help — Find what’s changed in a file
• Microsoft Support — View previous versions of Office files

Update log

September 29, 2026 — First publication.