What a PDF says about you before anyone reads it
Author, employer, software, the machine it came from, when it was written, when it was last touched. None of it is visible on the page, and all of it travels with the file.
Inspect and strip PDF metadataOpen a PDF you wrote last month, in any reader, and look at “Document properties”. You will usually find your name, the software that produced it, the software that last saved it, and two timestamps. Send that file to a client and you have sent all of it.
Where the data lives
A PDF can carry metadata in two places at once. The older one is the Info dictionary: a small set of key–value pairs with fields named Author, Creator, Producer, Title, Subject, Keywords, CreationDate and ModDate. The newer one is XMP, an XML block that can hold the same fields plus whatever an application decided to add — sometimes edit history, sometimes a document ID, sometimes a whole second metadata schema.
Both can be present. Tools that only clear the Info dictionary leave the XMP copy behind, and vice versa. This is why “metadata removed” and “metadata still there” can both be true about the same file, depending on where you look.
What each field gives away
- Author — a name, sometimes a full name, sometimes a username from the machine.
- Creator / Producer — the application chain, e.g. a word processor that produced it and a PDF library that converted it. Together these often identify the organisation.
- CreationDate / ModDate — when it was written and last edited, in the writer's timezone. Useful for establishing that a “final” version was edited after the meeting.
- Title / Subject / Keywords — frequently the internal draft title rather than the one on the cover.
- XMP — anything the authoring tool felt like recording, including the original file path.
None of this is a vulnerability in the PDF format. It is a feature for archivists, and a leak for everyone else. The same fields are how a whistleblower's document gets traced back to a printer.
What stripping does not do
Removing metadata does not change the words on the page, and it does not remove the text itself — the content stream still holds every sentence, selectable and searchable. If the document is a scan, the image is unchanged: the picture of your desk is still in the picture.
Nor does it remove what is drawn on the page. A footer with your company name, a header with the matter number, a signature block: those are content, not metadata, and they need redaction rather than cleaning.
The check worth doing before you send
- Open the file in a reader and read Document properties.
- If the fields are filled, decide which ones you are happy to publish.
- Strip the rest, then reopen the file and confirm the fields are empty — both the Info dictionary and XMP.
- Re-save a copy rather than overwriting the original, so you keep one for your own records.
Doing it in the browser
PDF Metadata Inspector reads both stores, shows you every field grouped by how much it reveals, and lets you blank them and download a clean copy. It runs in the tab, which is the right shape for this job: the file you are cleaning is, by definition, the one you do not want to hand to a stranger — and the metadata you are removing is exactly the kind of thing a server log would keep.