No tool can prove authorship from the finished text. A record of the writing process can, and most of that record already exists if you know where to look.
Last reviewed 12 November 2026
A detector score is a guess based on the final words on the page. A drafting history shows the work being made: the false starts, the paragraph that moved, the sentence rewritten four times. When a score and a history disagree, the history is the stronger evidence, and most reasonable reviewers treat it that way.
This guide covers two situations: recovering evidence for something you have already written, and setting up habits so the evidence is there next time.
Do these before you edit, move or rename anything. Some histories are lost when a file is copied or exported.
Take screenshots of the version list showing dates and times, then leave the original file where it is. A share link to the live document is better than an exported copy, because the reviewer can step through the history themselves.
Offer to discuss the work in person. Someone who wrote a piece can explain why an argument is ordered the way it is, what they cut, which source changed their mind and where they got stuck. That conversation is hard to fake and is often what settles the matter.
Do not rewrite the piece to get a lower score, and do not run it through a "humanizer". Either one alters the text under dispute, destroys the match between your history and your submission, and looks like an attempt to hide something.
None of these take extra time once they are set up.
A history with steady, uneven progress over days is strong evidence of ordinary writing. It is not mathematical proof, and a reviewer may still have questions. Equally, a missing history proves nothing: plenty of honest writers draft on paper, offline, or in tools that keep no record. If that is you, say so plainly and rely on your notes, your sources and a conversation about the work.
If you used AI in a way your course or publisher permits, say exactly what you did. A clear account of permitted use is far easier to defend than a blanket denial that later turns out to be incomplete.
Already facing an accusation? The step-by-step guide is Accused of using AI when you didn't. For why the score itself is weak evidence, see What a 1% false-positive rate actually means.