Remove metadata and tracked changes from Word documents

How to remove metadata in Word itself (Document Inspector)

Word has a built-in tool, and for many documents it is enough. Use it first:

1. Windows: File → Info → Check for Issues → Inspect Document.
2. Mac: Review → Protect → Protect Document, or Tools → Protect Document, then tick "Remove personal information from this file on save".
3. Tick the categories to inspect, click Inspect, then Remove All next to each finding.
4. Save the file. Work on a copy — several removals cannot be undone.

What Document Inspector does not remove

This is the part that catches people out. Microsoft's own documentation notes that the inspector cannot remove everything, and independent research has looked closely at what survives:

• Revision Save Identifiers (rsid). Word stamps editing sessions with rsid values throughout the XML. They are the subject of published forensic research — they can be used to work out how many sessions a document went through and to link documents edited on the same machine. Document Inspector does not target them.
• Metadata inside embedded images and charts. A pasted screenshot or photo carries its own EXIF data — camera, and sometimes GPS location. Inspecting the document does not reach inside embedded media.
• Unaccepted tracked changes. Deleted text remains in the file until changes are accepted or rejected. Hiding markup (Review → No Markup) only changes the display; the text is still in the XML.
• Anything the inspector declines to touch because removing it could break the document's layout — it reports those and leaves them.

The tool above works at the file level instead: it resolves tracked changes, deletes comments, strips rsid attributes, clears the author and company fields, and drops printer-settings parts — then tells you exactly what it found so nothing is silent.

See it yourself: unzip the document

A .docx is just a ZIP archive. Rename any copy to .zip, extract it, and look inside — no special software needed:

docProps/core.xml    → dc:creator, cp:lastModifiedBy, cp:lastPrinted
docProps/app.xml     → Company, Manager, template name, editing time
word/document.xml    → w:ins / w:del (tracked changes), w:rsid* attributes
word/comments.xml    → every comment, with its author
word/settings.xml    → rsid table for the whole document

Open docProps/core.xml in any text editor and your name will usually be the first thing you see. This is also how anyone receiving your document can read it — no forensic tooling required, which is the reason scrubbing matters.

Why lawyers and firms scrub metadata before sending

In legal practice, sending a draft with tracked changes and comments intact can expose negotiating positions, internal disagreement, or privileged commentary to the other side. University IT departments and bar association technology guidance both publish metadata removal instructions for exactly this reason. The same applies to anyone sending contracts, tenders, grant applications or client deliverables: the document you *see* and the document you *send* are not the same thing.

Anonymising a manuscript for double-blind peer review

Journals that use double-blind review ask for anonymised submissions, and author metadata is the most common way anonymity is accidentally broken — the reviewer opens the file properties and sees your name. Before submitting: run the file through the tool above, check that self-citations in the text are phrased in the third person, and confirm the filename itself does not carry your name.

Does removing metadata change the document?

Not visually. Tracked insertions are accepted and deletions dropped — the same outcome as Accept All Changes — comments are removed, and the metadata fields are cleared. Text, formatting, images, tables and layout are untouched. You always get a new file (with -clean added to the name) so your original stays exactly as it was.