← Back to AI Watermark Remover
Document metadata remover
Clear author, company and editing-time properties from DOCX, ODT, PDF and web files without changing a word of the body.
A DOCX is really a ZIP. Inside, docProps/core.xml holds author, last-modified-by, creation time and revision count, while docProps/app.xml holds company, manager, total editing time and template name. Almost nobody fills these in deliberately — they are inherited. The template you got from a colleague may still name them as author; the CV you send out may still carry your previous employer in the company field.
The failure modes are concrete: an author name left in a double-blind submission voids it; a tender document carries the previous client's name in its properties; a "carefully prepared" deliverable reports three minutes of total editing time. None of this requires technical skill to see — right-click, properties.
Cleaning also fixes the declarations that point at what was removed: dropping customXml also clears the matching Override in [Content_Types].xml and the Relationship entries in _rels, so Word isn't left with dangling references. For ODT, the mimetype entry stays first in the archive and uncompressed, which the ODF spec requires — get it wrong and the file won't open.
How to use
Drop in documents
DOCX, ODT, PDF, SVG, HTML, Markdown and plain text, several at a time.
Choose how hard to scrub
"Clear everything" empties all property fields; turning it off only clears values that look AI-written, keeping an author name you set yourself.
Download the clean copy
The log names each cleared field and its original value so you can verify.
FAQ
Is the body content modified?
No. word/document.xml and ODT's content.xml are copied verbatim, as is embedded binary content such as images. Only property parts and customXml change.
Are comments and tracked changes removed?
No. Those are body content rather than metadata, and removing them automatically is too risky — you may well need them. The report flags that a document contains comments; what to do about it is your call in Word.
How thorough is PDF cleaning?
It rewrites the document info dictionary (title, author, subject, keywords, creator, producer) and drops the XMP metadata stream, which covers the large majority of cases. PDF structure is complex, though, and traces inside embedded objects can survive; a pure-browser tool won't match a dedicated desktop one. Said plainly rather than glossed over.
Does it remove text hidden under a black box?
No — and this is a genuine source of leaks. Drawing a black rectangle on a PDF only covers the text graphically; it can still be copied out. The fix is to delete the text before generating the PDF, or use a real redaction feature.
What's the difference between the two modes?
Conservative mode only clears a field when its value mentions things like Claude, OpenAI or Gemini, keeping an author name you set. Thorough mode empties every listed property regardless of content. For a CV, use thorough.
Related searches
Other tools
AI text watermark & invisible character remover
Find and strip zero-width characters, tag characters, bidi controls and odd spaces. A visual preview lets you actually see what was invisible.
Image metadata & C2PA remover
Strip EXIF, XMP, C2PA content credentials and generator fields from PNG and JPEG, while optionally keeping the ICC profile so colours don't shift.
AI watermark detector
Read-only. Lists the metadata, C2PA manifests and AI traces a file actually contains, so you can decide before changing anything.