What your PDF says about you, and the awkward part of removing it

Export a document from Word and the PDF quietly records who made it, what software made it, and when. None of that appears on the page. Most people never look.

The information itself is well documented — there are plenty of articles listing the fields. The part that goes unmentioned is the awkward bit at the end of them: you're advised to remove this data from a confidential document, and the recommended way to do that is to upload the confidential document to someone's server.

What's actually in there

A PDF stores document information in a dictionary alongside the page content. The fields that matter:

Author is the one that catches people. When you export from Word, this field is usually inherited from your operating system account name or the name on your Office licence — not something you typed. An anonymous submission, a report published by an organisation, a resume sent to a recruiter: all of them commonly carry a name nobody meant to include.

Producer and Creator name the software, frequently down to the exact version. That tells a reader what you run, and tells anyone curious which known vulnerabilities to check against.

Creation date and modification date reconstruct a timeline. A contract whose modification timestamp is four minutes before it was signed says something the document itself doesn't.

Title, Subject and Keywords are usually harmless, occasionally not — a title left over from the template you reused can name a previous client.

You can see all of this without any special software. Open the file in most PDF readers and look at document properties — or load it into our metadata editor, which lists every field at once.

The circular problem

Here is the shape of nearly every guide on this topic: your PDF contains sensitive metadata, so use this tool to remove it. And the tool asks you to upload the file.

A document that matters enough to worry about its metadata is a strange thing to hand to a third party. You've uploaded the full contents in order to remove a few fields from it.

That's not a reason to distrust those services. Most delete uploads on a stated schedule and are transparent about it. But deletion after the fact isn't the same as the file never leaving, and for a document covered by an NDA or a workplace data policy, the difference is the whole point.

Metadata editing does not require a server. It's a change to a small dictionary in the file, not a re-interpretation of the document. Our metadata editor reads and writes all eight fields entirely in your browser — you can confirm that by disconnecting from the internet after the page loads and using it anyway.

Two things it can't reach, said plainly

We'd rather tell you the limits than let you assume a clean sweep.

The XMP stream. PDFs can carry a second metadata layer, an Adobe-defined XML block stored alongside the info dictionary. The JavaScript library this runs on exposes no way to touch it, so a file with XMP data will still have it after clearing. Most articles on this topic don't mention that a second layer exists at all.

EXIF inside embedded images. A photograph or scan placed into a PDF keeps its own metadata block, which on a phone photo can include GPS coordinates. Editing document metadata does nothing to it. Clean the images before they go in — the same applies whether you build the PDF from images or embed them in Word first.

Clearing metadata is not redaction

These get used interchangeably and they are completely different operations.

Metadata is information about the file. Redaction removes content from the page.

Covering text with a black rectangle changes what's displayed, not what's stored. The original text usually sits underneath, still selectable, still copyable. This has produced real disclosures in published legal filings. No metadata tool fixes it, and neither does exporting to PDF.

Text that needs to be gone has to be removed from the source document before the PDF is made, or with a tool built specifically for redaction.

One more consequence worth knowing: editing metadata invalidates an existing digital signature. The signature covers the file's contents, and you've changed them. Check signed documents before saving over them.

A detail from building this

While extending our own editor we found something worth passing on.

pdf-lib, the library most browser-based PDF tools are built on, writes its own name into the Producer and Creator fields and resets the modification date every time it saves a file. That's the default behaviour, and it has to be explicitly switched off. It applies to every operation that rewrites a PDF, not only metadata editing — merging, splitting and rotating all pass the file through the same save step.

Which means a metadata tool built on it, without that flag set, will clear your author name and then stamp a fresh software fingerprint and a new timestamp into the file. It hasn't anonymised the document. It has replaced one set of identifying data with another.

We found this in our own code and switched it off. It's worth checking in any browser-based PDF tool: clear the fields, save, and reopen the output's properties. A Producer field naming a PDF library means the tool wrote to your file on the way out.

The short version

Before sending a document you'd rather not be traced back to you:

1. Open its properties and read the Author and Producer fields. Most of the time that's the whole problem.
2. Clear what you don't want in there — locally, if the document is sensitive.
3. Reopen the output and confirm the fields came back empty, including Producer.
4. Remember that none of this touches the text on the page, or the EXIF in images you embedded.

Four steps, one minute, and it's the one document-hygiene habit that most reliably catches something you didn't mean to send. Worth doing before you merge a bundle for sending, since the merged output inherits fields from whichever file went in first.

If you want the wider picture of which PDF operations need a server at all, we wrote about where browser-based tools hit their limits, and the privacy page breaks this site down tool by tool.