Can a detector result be proof?
No. A detector result is evidence, not proof. It cannot tell a personal style from a machine one, and it errs in both directions. At school and in a newsroom it can start a conversation, but no sanction can be built on it.
Why it does not hold up as proof
A detector works with probability and has no access to what happened while the text was being written. False positives and false negatives are common with every tool.
The number also cannot be reproduced with a different tool — and what cannot be verified independently hardly carries the role of proof.
What holds up instead
The version history of the document, notes and source material, and the author's ability to talk about the content and defend the decisions in the text.
Verification questions about the content of the work are the strongest tool in practice. Whoever wrote the text answers without preparation.
How to hold that conversation
Don't start with a number. Start with the text — a specific passage and a specific question about it.
Leave room for an answer. Part of the suspicion dissolves as soon as the author explains where the passage came from.
FAQ
Can I fail a student on the basis of a detection?
Not on the basis of the number alone. The decision has to rest on the content of the work and on the conversation, not on the output of a tool.
What is detection for if it is not proof?
It shows where to look carefully. Across thirty papers that saves hours.
Should a school announce that it uses detection?
Yes. A clearly written rule set out in advance resolves most of the disputes that would otherwise arise.
Try it on your own text
Paste a text or upload a file and look at the score and at the specific sentences that came out suspicious.
Run a detection