Can the old metadata be recovered from the file afterwards?

Not from the output this produces, because the document is serialised again from its object graph rather than patched in place, so removed objects are simply never written. That is worth stating precisely: the guarantee covers the PDF's own metadata structures, and it does not extend to metadata inside embedded images, which this tool does not touch.

Why does the file size change after removing metadata?

Rewriting the document from scratch also repacks it, so the result is often slightly smaller than the original even beyond the bytes that were deleted. Nothing visible is re-encoded — the page content is carried across as it was.

Does it work on scanned PDFs?

Yes. Whether a document has a text layer has nothing to do with whether it carries metadata, and a scan usually carries plenty — the scanner software names itself in the producer field, and the creation date records when the page went through the machine. The scanned image itself keeps its own embedded metadata, which is the one limit worth repeating here.

Does the parsing block the page while it runs?

No. The document is parsed in a background thread, so the page stays responsive even while a large file is being read. That also means the library doing the parsing is never loaded into the page itself — it is fetched only once you actually add a PDF.

PDFRuns in your browser

PDF metadata remover

It shows you what the document is carrying first. Deleting is a second, deliberate press.

1Files

Drag PDFs in, paste, or click to choose

A whole folder works too · Ctrl/Cmd+V pastes · on phone this opens your photo library · up to 100 MB per file

2Inside

Add a PDF and everything it is carrying gets listed here first. Nothing is deleted until you press the button, and nothing is uploaded at any point.

3Remove

Mechanism

How this PDF metadata remover rewrites the file

A PDF is a graph of numbered objects, and its metadata sits in a handful of known places: the Info dictionary, an XMP block hanging off the catalogue, per-page XMP, private application scratch areas, and the authorship stamped on each comment. This tool parses the object graph, reports every one of those it finds, and on request removes both the reference and the object itself before writing the whole document out again from scratch. Writing from scratch is the part that matters. Editing tools commonly append their changes to the end of the file instead, which leaves the previous version of every object sitting in the bytes, readable by anything that follows the chain backwards. A full rewrite carries only the objects that survived.

Not taking our word for it is the point: the code that does this is on GitHub, tests included.

Limits

What it cannot do

Metadata inside embedded images is not touched. A photo placed in the document keeps its own EXIF, including any GPS coordinates, because that data lives inside the image stream rather than in the PDF's own structures.
Attachments, scripts and filled-in form values are reported but deliberately left in place. They are the document's content and its behaviour, not metadata, and a tool asked to strip metadata should not quietly delete them.
Encrypted PDFs are handed back untouched. Rewriting a document we cannot properly open would either break it or silently lose parts of it.
Text in the pages themselves is content, not metadata. If a name appears in the document body, it stays there — check the visible text separately.

Questions

Questions people actually ask

Can the old metadata be recovered from the file afterwards?: Not from the output this produces, because the document is serialised again from its object graph rather than patched in place, so removed objects are simply never written. That is worth stating precisely: the guarantee covers the PDF's own metadata structures, and it does not extend to metadata inside embedded images, which this tool does not touch.
Why does the file size change after removing metadata?: Rewriting the document from scratch also repacks it, so the result is often slightly smaller than the original even beyond the bytes that were deleted. Nothing visible is re-encoded — the page content is carried across as it was.
Does it work on scanned PDFs?: Yes. Whether a document has a text layer has nothing to do with whether it carries metadata, and a scan usually carries plenty — the scanner software names itself in the producer field, and the creation date records when the page went through the machine. The scanned image itself keeps its own embedded metadata, which is the one limit worth repeating here.
Does the parsing block the page while it runs?: No. The document is parsed in a background thread, so the page stays responsive even while a large file is being read. That also means the library doing the parsing is never loaded into the page itself — it is fetched only once you actually add a PDF.