




the real and only unchatgpt
Remove Claude watermark from AI text. A full rewrite strips the invisible Claude text mark. Paste Claude output and humanize it.
Works with your favourite AI tools:
The mark is in word choice, not hidden unicode. Rewrite the whole draft, then copy it out.
Drop in output from Claude.ai, Claude Code, or the API. Light edits can keep the mark.
The Claude watermark lives in word choice. A full rewrite picks new words, so the old pattern is gone.
Use the rewritten draft. Hidden-character cleaners do not remove this mark.
Claude does not add hidden characters. New models change which next word they pick, so Anthropic can later score whether Claude wrote some of the text.
It nudges sampling when the model has a real word choice. Exact code tokens often skip that nudge. Prose carries more of the mark than code.
Rewrite the text so Claude no longer chose the words. Paste it in the form above. Stripping zero-width spaces does not remove this mark.
Often yes. Proofreading leaves most of the original word pattern. A complete rewrite is what removes it.
Less than prose. Comments can still carry it. Do not treat this tool as a code-strip utility.
No public text checker exists. Anthropic keeps the Detection API private. Their file checker only reads C2PA metadata on images and media, not prose.
Models that launched on or after 2 August 2026 mark text from day one. Older models get the EU delay until 2 December 2026.
Unicode cleaners on GitHub only strip invisible characters. That is not the Anthropic mark. This page rewrites the text instead.
Same tool, broader copy on the UnchatGPT home page.