DOCX · OOXML · document properties
MarkdownA Word file stores more than the words. Clean the package, not the document body.
A DOCX file is a ZIP package containing the document plus properties, revision identifiers, comment authors, a preview thumbnail, and sometimes metadata inside embedded images. Gifi removes those named metadata surfaces while preserving the document body and title. It refuses a package it cannot clean completely.
Reviewed by Can Balkaya, CTO · Published
Process
How to verify the job
01
Inspect the DOCX
Drop the original file into Gifi to preview the package metadata before it is uploaded.
02
Clean the package
Continue to remove document properties, revision IDs, comment authors, the thumbnail, and metadata inside supported embedded images.
03
Verify the download
Reinspect the cleaned DOCX and compare the named actions; the paragraphs, tables, styles, and title should remain.
Scope
Supported and unsupported surfaces
| Surface | Gifi action | Verification limit |
|---|---|---|
| Named metadata or invisible Unicode | Inspect, remove, and report named actions | Reinspect the returned bytes independently |
| Statistical patterns in ordinary words | Rewrite the prose | No certificate against a private detector |
| Visible logos, pixels, audio, or video | Unsupported | No removal claim |
What Gifi removes
The cleaner clears author and company fields, custom properties, revision-save identifiers, comment authors, and the page-one preview thumbnail. It also opens supported images embedded in the package and removes their EXIF, XMP, comments, and local C2PA containers.
Every removal is added to the action list returned with the cleaned file. That list is evidence about named package parts, not a general claim that the document was never edited or AI-assisted.
What stays in the file
The document body, tables, styles, links, and title stay. Gifi rebuilds the ZIP package after changing the metadata parts; it does not convert the document to another format or rewrite the prose.
Macros and unsupported embedded formats are not silently half-cleaned. If the package cannot be handled safely, the job is refused and the credit is refunded.
A concrete verification path
Inspect the original in the browser, save the list of findings, then clean and inspect the returned file again. For a deeper check, rename a copy from .docx to .zip and compare docProps, comment records, settings, and the thumbnail folder.
This verifies the package surfaces Gifi names. It does not remove a watermark printed into an embedded picture, and it does not certify the output against a private text detector.
Questions this usually raises
- Does removing DOCX metadata change the document text?
- No. The body, tables, styles, links, and title are preserved. The cleaner changes metadata parts and package identifiers.
- Are comments deleted?
- The comment text can remain, but author identity in comment records is cleared. Review the result if comments themselves are sensitive.
- Does this remove visible watermarks in Word?
- No. A visible header, background image, or mark printed into an embedded image is document content, not metadata.