PARTLY

Partly true: Claude really does watermark its text and the free remover really strips metadata and hidden characters, but the word-choice watermark only breaks with a real rewrite, the 'Claude Rewriter' button actually rewrites with ChatGPT, and ChatGPT's own watermark is EU-only so far

Facebook · Oct 2026

The claim'Claude and ChatGPT are leaving watermarks on text,' so even if you write your own ideas and only use Claude or ChatGPT to clean up, 'you're going to be flagged as AI-generated text.' Here is 'a free tool to kill that watermark': it removes obvious metadata from files and removes text-based watermarks; 'if you're using Claude, make sure you turn on Claude Rewriter.' 'No guarantee this is 100%.' Follow, comment WATERMARK and share for the link.

The premise is half right. Anthropic's own help center confirms an imperceptible watermark woven into all Claude-generated text, with no opt-out, plus signed Content Credentials on Claude-made image files. ChatGPT is different: OpenAI announced on Oct 5, 2026 that it is adding a text watermark to ChatGPT and Codex in the EU only, with API watermarking off unless a developer turns it on. So 'ChatGPT watermarks your text' is not true for most US users today. The tool is real and free with no sign-up, at watermark.sabrina.dev. It strips invisible characters, odd spacing, and C2PA, EXIF and XMP metadata from files, which works and which any metadata cleaner also does. But Claude's text mark lives in word choices, not hidden characters, so only a real rewrite touches it. The toggle labeled 'Claude Rewriter' says on the page that it rewrites with ChatGPT, not Claude. That avoids re-marking by Claude, but it means your polished draft is now ChatGPT's wording. 'Flagged as AI' is also overstated: Anthropic's text detector is in private preview for regulators, media and researchers, so ordinary checkers cannot read the mark; what flags people is third-party detectors judging style. Credit where due: she says plainly there is no 100 percent guarantee. The ask is follow, comment and share, and the tool page funnels to her free playbooks email list and her company Blotato.

What holds up

Reel transcribed and all 16 contact-sheet frames viewed; on-screen tool is watermark.sabrina.dev with 'Claude Rewriter' and 'Choose a file' options.
Tool page read (Oct 30, 2026): Layer A hidden characters, file metadata removal (C2PA, EXIF, XMP), Layer B rewrite; Rewriter 'rewrites with ChatGPT, then cleans hidden text marks'; 'No sign-up'; links to Blotato and an email playbook signup.
Anthropic help center article 16266773 read: watermark applies to all generated text on every surface; heavy editing, paraphrasing or translation may remove it; text detection API is private preview only.
TechCrunch, Oct 5, 2026: OpenAI text watermark rolls out to ChatGPT and Codex in the EU only; API watermarking is opt-in and off by default.
x-ref #181 and #271 (Claude watermark removers, both bs); x-ref #272 (same creator's Blotato-linked skill).

What doesn’t

'ChatGPT watermarks your text' is EU-only as of Oct 2026.
The 'Claude Rewriter' toggle rewrites with ChatGPT, so your words get replaced either way.
'You will be flagged' overstates it; Anthropic's text detector is not public.
Follow, comment and share gate plus links to the creator's own company, Blotato.

The catch

Metadata stripping is real and easy. The word-choice watermark only goes away when the words change, and this tool changes them by handing your draft to a different AI.

How to actually do it

If you only need file metadata gone, re-save or export the file, or use any free EXIF or metadata cleaner.
Keep your own wording: use AI for suggestions, then make the edits yourself rather than pasting back its full rewrite.
Where a school, client or publisher requires AI disclosure, disclose; no remover changes that obligation.
Use the free tool at watermark.sabrina.dev directly if you want it; no comment or follow is needed to reach it.

A working free cleaner with an honest no-guarantee line, sold on a premise that is only half current. The real fix for 'my edited draft reads as AI' is doing more of the editing yourself.

Confidence: Medium
Posted by: Sabrina Ramonov, an AI creator and founder of the Blotato social media tool (about 152,000 plays); reel sent to Buddy Mon Oct 26, 2026, ~6:58 AM

See the original claim →

We test hype for free. We build the real thing for a living.

Thirty minutes, no pitch — and you'll leave with something useful either way.

Book a call with Todd or start with the free Business Checkup →

The Verdict Weekly

Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.