What a PDF reveals about you
A document you send carries more than the words on the page: who wrote it, with what software, when, and sometimes a full edit history. Here is what is in there and what to do about it.
Last updated: 20 November 2026
The four layers of hidden information
It helps to separate them, because they need different tools and carry different risks:
- Document information — title, author, subject, keywords, producing software, creation and modification dates.
- XMP metadata — an XML packet that often holds much more: edit history, application versions, colour profiles, camera details, sometimes the original author and the last editor.
- Content-level traces — author names in comments and tracked changes, default values in unfilled form fields, filenames referenced by embedded objects.
- Structural traces — incremental updates that leave previous versions of the text inside the file, and text covered by a drawn rectangle rather than removed.
Layer 1 and 2: the part you can fix in one click
Open any PDF in a reader and check File → Properties. You will usually see a name you did not type — the account name of whoever exported the document. Word writes the Windows profile name; LibreOffice writes its own; scanning software writes the device model. None of it appears on the page, and all of it travels with the file.
XMP is the same idea in XML, and it is worse behaved: many readers never display it, yet it is plain text inside the file and any script can dump it from an entire folder in seconds. If you have ever received a document whose "author" was a colleague's laptop account, that is what you were seeing.
Both can be removed without touching the page contents — that is what the metadata tool does, and it verifies the result by re-reading the output rather than assuming success.
Layer 3: comments, forms and the things metadata cleaning misses
Comment author names are page content, not file metadata. Removing them either breaks the annotation structure or leaves the comment orphaned, so a metadata cleaner is right to leave them alone. If comments contain notes you would not send, delete the comments themselves in a reader before exporting.
Unfilled form fields are subtler. Many PDFs carry default values from the template — a company name, a branch code, sometimes an email address — that never display because the field looks empty. Flatten the form, or remove the fields, if the file leaves your organisation.
Layer 4: the fake redaction problem
Drawing a black rectangle over text does not delete the text. It draws a black rectangle on top of it, in the same content stream. The characters are still there: searchable, selectable, and extractable in under a second with the right tool. This has caused real, documented disclosures — in court filings, in government releases, in corporate reports.
Proper redaction means removing the underlying text and then removing the objects that contained it, then verifying by extracting text from the result. Acrobat's Redact tool does this. Drawing shapes does not, and neither does any converter.
The other structural trace is incremental update: a PDF can be saved by appending changes rather than rewriting the file, so an earlier version of a sentence — or an earlier version of a whole page — can survive inside the file after you "changed" it. A linearized or fully rewritten export removes them; simply saving in some editors does not.
A practical checklist before sending
- Clear document information and XMP, then verify — re-open the output and look at Properties again.
- Select all the text with
Ctrl/Cmd+Aand copy it into a plain editor. Anything you did not expect to find is a leak. - Search the document for terms that should not be there — the client you dropped, the old version number, the internal codename.
- Check for comments and tracked changes; delete, do not hide.
- If you redacted anything, extract the text afterwards and confirm the words are gone.
- If the file is signed and the signature matters, stop — any of the above invalidates it.
What "cleaned" does not mean
Removing metadata is not anonymisation. The content still identifies people, the writing style still identifies the author, and the file may still carry watermarks, tracking identifiers or embedded fonts that hint at the source. For documents going to court, to a regulator, or into a public disclosure, use a purpose-built redaction workflow and have someone check the output — a browser utility is the wrong level of assurance for that job.
Related tools
- Remove metadata from a PDF — layers 1 and 2, with a read-back check.
- Extract text — the fastest way to see what is really inside a document.
- Delete pages — when a page should not be in the file at all.