Cleaning tool · PDF
Clean PDF metadata
A PDF carries metadata in two places: the Document Info dictionary and the XMP stream. Both expose the author, the software and the creation time — even when nothing of the sort appears in the text.
Document Info/XMP stream/15 MB per file/10 per batch
Drop a file here, or click to choose
PDF only · 15 MB per file · up to 10 per batch
You can also paste a screenshot with ⌘V / Ctrl+V
Cleaning options
Hidden metadata only: visible text, images, fonts, bookmarks and form fields on the page are not modified. That also means if what you need to hide is the body content, you want a redaction tool, not a metadata tool.
The two metadata locations in a PDF
| Location | Contents | Action |
|---|---|---|
| Document Info dictionary | Title, Author, Subject, Keywords, Creator, Producer | Fields emptied |
| Dates | CreationDate, ModDate | Removed as well |
| XMP stream | An XML packet, possibly holding workflow history and AI markers | Stream object deleted |
| Catalog reference | The /Metadata entry pointing at the XMP stream | Deleted too, or readers keep looking |
| Page content | Text, images, vector art | Untouched |
Why cleaning a PDF means rebuilding the file
PDF is not a “tags plus pixels” format. It is a set of objects, with a cross-reference table (xref) recording the byte position of each one. Delete an object or shorten a dictionary and every later object shifts — update the xref without accounting for that and the file will not open.
So the approach here is: parse every object, rewrite the two locations being cleaned, then regenerate the xref table and the trailer. The resulting structure is self-consistent, and any standard reader can open it.
Object streams (ObjStm) are expanded into independent objects — otherwise the rebuilt traditional xref cannot reference compressed objects, and some readers will throw errors.
Common cases
- Resumes: the Author field in Document Info is still the name of whoever built the template.
- Quotes and contracts: Producer names an internal system, exposing your toolchain.
- Portfolio PDFs: Lightroom's export writes its own name and version straight into Producer.
- Decks exported from Canva or Figma: the Author field is often your account email.
- Scanned documents: ModDate reveals when the file was last edited, which matters for blind review.
Common questions
Does cleaning change how the PDF looks?
No. Only the hidden metadata dictionary and the XMP stream change. Text, images, layout and bookmarks are all unchanged.
Can it handle encrypted PDFs?
A password-protected PDF usually cannot be parsed. Remove the password in a reader first, then upload it.
Can it delete metadata inside images embedded in the PDF?
No. Embedded images are independent binary objects; this tool neither extracts nor rewrites them. Export them as JPG/PNG and clean those separately.
Does the file get smaller?
Usually slightly, because the XMP stream is removed whole. But the bulk of a PDF is fonts and content streams, so the change is rarely noticeable.
Why does a new date appear after cleaning?
Some readers write a fresh ModDate when they open and re-save a file. That is the reader's behaviour, not information carried over from your original.