Own your words
Remove AI watermarks and metadata from text and files
Result
Frequently asked questions
Is it legal to remove watermarks?
On content you own, yes, for privacy, hygiene, or workflow reasons. It is not legitimate for academic fraud, impersonation, or faking human authorship. Transparency laws (EU AI Act Article 50, California SB 942) may still require you to disclose AI origin when you publish.
Does it work on ChatGPT, Claude and Gemini text?
Invisible Unicode marks (zero-width characters, bidi, homoglyphs) are removed losslessly for any provider. Statistical token watermarks, the kind Google's SynthID-Text deploys, are not removed here: that requires rewriting the wording, which this tool does not do. No tool can honestly certify that a vendor detector will fail, because vendors do not publish detection keys.
Which file formats are supported?
PNG, JPEG, SVG, PDF, DOCX, ODT, HTML and Markdown get full container/metadata cleaning; plain text (TXT, JSON, CSV, code files) gets the Unicode scrub. PDF cleaning is strongest when exiftool is installed on the server.
Is my content stored or shared?
No. Text is processed in memory and never persisted. Uploaded files go to a temporary scratch area, get cleaned, and the cleaned copy is deleted as soon as you download it. We store only your email (for the daily quota) and a timestamp, never your content. We do not sell or share data.
Does this remove SynthID from images?
Pixel-level watermarks (SynthID images, video, audio) cannot be removed by third parties. That needs the original generator. What we do remove is the metadata layer (C2PA manifests, EXIF, XMP), so the file no longer advertises its provenance in its properties.
How do I remove a ChatGPT, Claude or Gemini watermark from text?
Paste the text into the tool and hit Clean. The Unicode scrub removes invisible characters (zero-width spaces, bidi controls, homoglyphs) losslessly, and it works for any provider — ChatGPT, Claude, Gemini and open-source models alike. Statistical token watermarks, like Google's SynthID-Text, are not removed: that needs rewriting the wording, which this tool does not do.
Can you remove SynthID watermarks?
Partially. SynthID image, video and audio marks are embedded in the content itself and cannot be removed by third parties. The metadata layer around them (C2PA, EXIF, XMP) is always removable. SynthID-Text (Gemini text) is a statistical signal spread across the prose; removing it means rewriting the wording, which this tool does not do.
How do I strip C2PA metadata from a photo?
Drop the PNG or JPEG into the tool and download the cleaned copy. The C2PA manifest, EXIF and XMP blocks are stripped deterministically — the pixels are untouched. Verify the result with c2patool or Content Credentials Verify, which will confirm no provenance manifest remains.
Is there a free AI watermark remover?
Yes — this is one. Watermarks Remover is free, with no advertising cookies and no content retention. Traffic is measured with Google Analytics 4. Paste text or drop a file without an account; results unlock behind a free sign-in (Google or email) that gives you 5,000 words of cleaning per day, every day.
Does removing a watermark change my text?
No. Removing invisible Unicode characters and space homoglyphs is lossless — the visible text is byte-identical. Metadata stripping on files never touches pixels, text or formatting.
Disclaimer
Watermarks Remover is provided for privacy and hygiene on content you own and are legally entitled to process. It is your responsibility to comply with all applicable laws, academic policies, workplace rules and platform terms, including AI-transparency obligations such as the EU AI Act (Article 50) and California SB 942, and to verify the terms of any content before processing it. By using this tool you agree to these terms; you are solely responsible for how you use the results, and we disclaim all liability for misuse, including academic fraud, copyright infringement or misrepresentation of AI-generated work. See the full Disclaimer in the Terms of Service.