unmarkclaude.io

Free checker, no account

Claude watermark detector and checker: test any AI text watermark remover on a marked sample

Take a text carrying a watermark of the same class Claude and Gemini put in their output, run it through the remover you are weighing up, paste back what came out. Our detector reads the returned text and says whether the mark is still there, with the score and the date. The sample is also the closest thing to a Claude watermark example you can hold: a thousand words of ordinary prose with the mark inside the word choice.

How to detect a Claude watermark after a remover: four steps

  1. 1. Take the marked sample

    Around a thousand words of ordinary English prose carrying a watermark of the same class as the marks in Claude and Gemini output. Everyone gets the same text, which is what makes two runs comparable: the only thing that differs between them is the tool.

  2. 2. Put it through whatever you want measured

    Any AI text watermark remover: a paid humanizer, a free one, a character cleaner, a script of your own, a model you host yourself, whichever mark it says it takes out. We do not need to know which, and nothing about the tool is sent to us unless you choose to name it.

  3. 3. Paste back what came out

    Free. No email.

  4. 4. Read the answer

    Four answers: the mark is still there, the mark is not found, the score sits in the grey band, or the text you pasted cannot be tied to the sample. Under the answer stand the five figures of the run.

How to read the answer

The answer What it means
Mark found Our detector still finds the mark in the text you pasted. The tool did not reach it.
Mark not found Our detector does not find the mark in the text you pasted. On this sample, on this run, the tool reached it.
Grey band The score sits between the two. It is reported as it is, not rounded to either side.

Recent runs

Every run below was made the same way, on the same sample, by us or by the person who sent it in. The figures are ours in both cases: our detector measured them on the text that came back. Latest run 5 September 2026.

  1. 2026-09-05 guillaumemeyer/watermarks-remover, rewrite layer mark not found our run
  2. 2026-09-05 guillaumemeyer/watermarks-remover, character cleaning mark still found our run
  3. 2026-09-05 claudewatermark.com mark still found our run
  4. 2026-09-05 claudewatermarkremover.app mark still found our run
  5. 2026-09-05 favtutor.com mark not found our run
  6. 2026-09-05 gptcleanup.com mark still found our run
  7. 2026-09-05 unmarkclaude.io mark not found our run
  8. 2026-09-05 guillaumemeyer/watermarks-remover, rewrite layer mark grey band our run
  9. 2026-09-05 guillaumemeyer/watermarks-remover, character cleaning mark still found our run
  10. 2026-09-05 claudewatermark.com mark still found our run

Removers ranked by the mark they name

Claude watermark removers

7 tools that say they take the Claude watermark out of text, on the same marked sample. The first three:

Place Tool Mark Facts kept
1 unmarkclaude.io (ours) 0.8 not found 3 of 3
2 favtutor.com -1.0 not found 1 of 3
3 guillaumemeyer/watermarks-remover, rewrite layer -1.1 not found 1 of 3

The full ranking of Claude watermark removers, with the date of every run.

Frequently asked

How do I test a Claude watermark remover?

Take the marked sample above, put it through the remover, paste back what came out. Our detector reads the returned text and answers in a few seconds: mark found, mark not found or grey band, with the score and the date.

Which tools can I test?

Any of them. A Claude watermark remover, a Gemini one, a general humanizer, a character cleaner, a script of your own or a model you host. The sample carries a watermark of the same class as the marks in Claude and Gemini output, so every tool gets the same kind of answer.

Can I publish the result?

Yes. Name the tool and its address under the figures, and the run goes on that tool page after a look from us. The figures are never edited: our detector measured them on the text you pasted.