What Is PDF Metadata?
Document properties are the obvious part. The revisions left inside the file are the part that leaks.
Three places, not one
Most people meet PDF metadata through the Document Properties dialog, which shows a handful of fields: title, author, subject, keywords, the application that created the content, the library that produced the file, and two timestamps. That is the document information dictionary, and it is the easy part.
Adobe tools also write an XMP packet — an XML block attached to the document catalogue that duplicates much of the same information and adds identifiers that persist across exports, so two PDFs generated from the same source can be linked to each other.
The third place is not a field at all, and it is where the real disclosures come from.
How incremental updates preserve your edit history
PDF was designed so a file could be modified without rewriting it. When an editor saves a change, it can append only the objects that changed, followed by a new cross-reference table pointing at them. Readers use the newest table, so they show the current state of the document.
The previous bytes are still there. Nothing overwrites them; the file simply grew. A PDF edited five times can contain five generations of its own content, all readable to anyone who looks past the last cross-reference table.
This is why clearing the author field in an editor and saving is not enough. The properties dialog will show an empty author, and the original one will still be sitting in the file a few kilobytes earlier.
The failure this causes, repeatedly
The pattern behind most published PDF leaks is the same: a document is drafted, edited, partially sanitised, then published without ever being rewritten. The published file carries the drafts.
The same mechanism defeats naive redaction. Drawing a black rectangle over text adds a shape to the page; it does not remove the text underneath, which stays selectable and copyable. Combine that with an incremental save and a document can leak both its redacted content and its earlier drafts.
What actually removes it
The reliable operation is a full rewrite: parse the document's current object graph, drop the metadata, and serialise a fresh file with a single cross-reference table. Objects belonging only to superseded revisions are not reachable from the current graph, so they are not written out.
You can check this yourself on any cleaned file. Count the occurrences of the string `startxref` — one per revision. A properly rewritten PDF has exactly one.
Command-line tools do this too: `qpdf --linearize in.pdf out.pdf` rewrites the file, and `exiftool -all= in.pdf` clears the metadata but appends an incremental update, so the two are usually combined rather than used alone.
What a metadata cleaner cannot do for you
Removing metadata is not redaction. If sensitive text is present on the page — visible, or hidden under a shape, or in a layer that is switched off — it is content, and it stays. Proper redaction removes the underlying objects and needs a tool built for it.
It also cannot help with an embedded attachment, a comment thread, or an image inside the PDF carrying its own EXIF. Those are separate problems with separate answers.
Check what your PDF is carrying
PDFs record who made them, with what software and when — and often keep earlier drafts inside the same file. See what yours contains, then remove it.
Open Remove Metadata From a PDFFrequently asked questions
How do I see a PDF's metadata?
Document Properties in Acrobat, the Inspector in macOS Preview, or a metadata tool. To see whether the file has multiple revisions, count the `startxref` markers — a clean file has one.
Does removing PDF metadata change the pages?
No. Pages, text, fonts and images are copied across unchanged. The document renders and prints exactly as it did.
Why did my PDF get smaller after cleaning?
Usually because superseded revisions were dropped. A file that has been edited several times can shrink noticeably when it is rewritten.