You polished the cover letter, trimmed it to one page, and addressed it to the hiring manager by name. The file has one more thing to say, though, and it isn't on the page: [email protected], revision 23, 94 minutes of editing, a template named "Layoff_Memo_v3.dotm". Here is what's actually inside a Word file, and how to get it out.
Since 2007, Word files (and Excel and PowerPoint files) have been ZIP packages rather than single blobs. That matters because the package has rooms, and the metadata lives in its own rooms, separate from your text. docProps/core.xml holds the core properties: the author (pulled from the Microsoft account that saved it), last modified by, exact created and modified timestamps, revision number, and the title/subject/keywords block. docProps/app.xml holds the extended ones: Company (set when Office was installed on that machine, so it follows every file that machine produces), Manager, the source template, total editing time in minutes, and a hyperlink base. docProps/custom.xml holds custom properties, which in corporate environments often include sensitivity labels and internal identifiers. And an optional docProps/thumbnail.jpeg is a preview image that can show deleted content from earlier versions.
How much is that in practice? In a test we ran, a routine HR performance-review document carried 18 separate identifying items: nine core fields (author Jane Q. Author, last modified by, two timestamps, revision 7, title, subject, keywords, description), five app fields including the legal department's company name and a manager's name, three custom properties (a sensitivity label, a client name, a reviewer), and a thumbnail. The person forwarding that file had no idea any of it was riding along.
Most recipients never open the properties, but the exposure points are common enough. Windows File Explorer shows author and title in the preview pane and sorts by them. Google Drive and Dropbox index author fields into search. macOS Finder shows them under Get Info. Email attachments keep them intact end to end. And anyone who knows to look can open the details in two clicks. The table below maps each leak to where it shows up:
| Field | Where it comes from | Visible to recipients via |
|---|---|---|
| Author | The Office/Microsoft account that saved the file | Explorer details pane, Drive search, Get Info |
| Last modified by | Whoever saved it last (often a reviewer) | Properties panel, review tools |
| Company & Manager | Office installation defaults on that machine | Properties, document review software |
| Revision number / edit time | Word counts every save | Properties (94 minutes says "rewrote this five times") |
| Template name | The .dotm/.dotx the doc started from | Properties ("Layoff_Memo_v3" is a real problem) |
| Custom properties | Sensitivity labels, workflow tools | Properties, e-discovery and DLP scanners |
| Thumbnail | Saved preview image | Some file managers; can show deleted text |
Word has a native tool: File > Info > Check for Issues > Inspect Document. Tick the categories and Remove All. It genuinely removes document properties, personal information, headers/footers (if you ask), hidden text, and comments. Three caveats. First, results vary by Word version, and some users report fields reappearing after the next save if their account auto-fills author. Second, it operates one file at a time, which is painful for a folder of attachments. Third, the checkbox list tempts people into removing things like headers they actually wanted. Verify afterwards in the same Properties pane, because "Inspect says removed" and "properties are empty" are not always the same thing.
The other approach skips Word entirely. Because a .docx is a ZIP, you can unzip it, replace docProps/core.xml and docProps/app.xml with minimal versions containing no personal fields, delete custom.xml and the thumbnail, fix the manifest so nothing points at the deleted parts, and rezip. The document content is untouched, byte for byte. In the test above, the 8,185-byte HR document came out at 7,262 bytes with all 18 items gone and the body text intact. This is exactly what our Document Metadata Remover does in your browser (no upload, works on batches), and it's also how you can verify any cleaner: run it, then open the properties and check for yourself.
"I'll just export it to PDF" is the most common wrong move. A PDF has its own metadata: the Info dictionary (author, title, subject, keywords, creator, producer) and an XMP block that echoes the same fields. When Word exports, the author and title flow straight across. The clean-and-export order matters too: clean the .docx first if you need both formats, because cleaning one does not clean the other. Our remover handles the PDF side as well, emptying the Info dictionary and stripping the XMP stream; an 8-field test PDF came out with zero keys and no producer fingerprint.
Be clear-eyed about the boundary. Removing metadata does not remove tracked changes, comments, or hidden text: those are content, stored in the document body, and a metadata scrubber that touched them would be mangling your file. Accept or reject revisions and delete comments before sending, or run the Inspector's content checks. It also does not anonymize the text itself: a letter you signed is still a letter you signed. And it does nothing about files already sent; it protects the next copy out the door.
For your own documents, cleaning metadata is ordinary hygiene. One honest footnote: in litigation, documents can fall under evidence-preservation duties, and metadata itself can be discoverable material, so scrubbing files that are or may be subject to a legal hold is a decision for your attorney, not a website (nothing here is legal advice). Outside that narrow context, there is no downside to sending a clean file.
Open a .docx, .xlsx, .pptx or PDF and hit Clean. Author, company, history and custom properties are stripped locally — the file never leaves your machine.
Document Metadata Remover →Word files gossip: author, company, edit counts, templates, timestamps, all in parts of the package nobody looks at. Inspect Document gets most of it; a package-level clean gets all of it, verifiably, without Word installed. Exporting to PDF without cleaning just moves the gossip to a second file format. Strip the properties before the file leaves your machine, keep content edits separate from metadata scrubs, and verify the result in the properties pane before you hit send. For images you're publishing, the same logic applies to EXIF: our Image Metadata Remover handles photos, and EXIF Viewer shows you what a photo is carrying before you post it.