How to remove metadata from a PDF before publishing it

The page says nothing. The file says who wrote it, on which machine, with which template, when it was started, and when it was last edited at 2 a.m. Journalists, opposing lawyers and curious readers check. Before a document goes public, the labels should say only what you intend.

How to remove metadata from a PDF

1. Open Edit PDF metadata and choose the PDF. It is read in your browser, not uploaded, and the tool shows the Title, Author, Creator and Producer the file holds, and how many fields its XMP packet carries.
2. Tick Remove all metadata.
3. Select Save metadata. The document properties are deleted — Title, Author, Subject, Keywords, Creator, Producer, both dates and any custom fields — along with every XMP packet attached to the document, its pages or its images, and private authoring data. The file identifier is replaced with a random one, so the copy cannot be matched to the original by it.
4. Read the report. The saved copy is re-opened and searched for an information dictionary or XMP packet left behind; if one had survived (a photo can carry its own XMP packet inside the image data, which is not removed), you would get an error instead of a file.
5. Check the saved copy. Open it in Edit PDF metadata again: Title, Author, Creator and Producer should say Not set and XMP should say None. A viewer’s Properties panel may show only one of the two places metadata is kept, so it is not proof on its own; if you have ExifTool, exiftool -a -G1 clean.pdf lists what is left in both.

To change values rather than remove them, type a new Title, Author or Subject instead of ticking the box. That rewrites those three fields and removes the document’s XMP packet, so its old copies cannot contradict the new values; Keywords, Creator, Producer and the dates stay as they were, and packets attached to individual pages or images are left alone.

To publish with nothing but a title, do both: remove everything first, then open the cleaned copy, untick Remove all metadata, type the Title, clear the Author and Subject boxes unless you want them, and save again.

Should you remove metadata before publishing?

Yes, in most cases: check it, and remove whatever you did not choose to publish. Metadata never appears on a page, so it is easy to approve a document without ever reading it, and it goes out with every copy.

It matters most when:

• The Author is a person who should not be named — the account the software was registered to, a contractor, a colleague who has since left.
• Title or Subject carries old text — a previous draft’s heading, an internal project name, another client’s name left in a template.
• The document is sensitive or submitted anonymously. Remove everything, and treat metadata as only one of the places that can identify you (see Limits, below).
• The dates or software say more than you want — when the work started, how long it sat, which application and version made it.

Some of it is worth keeping on purpose:

• A Title you chose. Viewers use it. As of September 2026, the source code of PDF.js, the viewer built into Firefox, puts a PDF’s title in the browser tab ahead of the filename, taking the XMP copy before the document properties. That is also why a stale draft title is so visible.
• An author, when the credit is deliberate — an organisation’s name on its own report, for example.
• PDF/A structure. If the file must be PDF/A, clean the metadata first and convert last. PDF/A requires an XMP packet, so removing metadata ends conformance; Convert PDF to PDF/A writes a new packet built from the document properties that remain.

What is in there

A PDF’s descriptive metadata includes:

• Title, Subject, Keywords — usually left as whatever the source application defaulted to, which is often the first line of an old draft.
• Author — the name in the software’s licence or account, not necessarily the person who should be credited.
• Creator and Producer — the application that made the document and the library that wrote the PDF. They reveal software, versions, and sometimes a corporate template name.
• Creation and modification dates — including time zones, which hint at where a document was made.

And it lives in two places:

1. The Info dictionary, a small set of key–value pairs that has been part of PDF since the beginning.
2. The XMP packet, an XML document embedded in the file, which carries the same fields and can carry many more — document history, template identifiers, the identifier of the original file a PDF was derived from.

They are supposed to agree. Frequently they do not, because the tool that edited one did not know about the other. A document whose Info dictionary is blank can still announce its author in XMP, and some readers, PDF.js among them, read the XMP title first.

Clearing a field is not always deleting it, either. A PDF can be saved as an incremental update, which appends the change and keeps the earlier revision in the file for anyone who looks. ExifTool’s documentation says its PDF edits work this way and that “old information is never actually deleted from the file”, and recommends rewriting the file afterwards with qpdf --linearize in.pdf out.pdf. Edit PDF metadata always writes the whole file again instead of appending to it, and after Remove all metadata it checks the saved copy for a second revision and refuses to hand it over if there is one.

What is not metadata, but leaks anyway

• Bookmarks that quote internal headings.
• Comments and sticky notes, including their author fields.
• Form-field values and hidden fields.
• Attached files.
• Text under a black box — see How to redact a PDF properly.
• The filename.

Removing metadata touches none of these; each needs its own step.

Before you publish: the full order

1. Rename the file to something neutral.
2. If the document has comments or form fields, Flatten PDF — after deleting any comments that should not ship, since flattening preserves them visibly.
3. If any content must go, Redact PDF it now. Redact PDF builds a new file without the original’s document properties, but it writes its own — the name of the PDF library and the time the file was made — so clean the metadata after redacting, not before.
4. Remove the metadata as above. If you want a Title, add it to the cleaned copy in a second pass, as described there.
5. Open the result in Edit PDF metadata again and check that XMP says None. Then, in a viewer, check the bookmarks panel and the attachments panel.

If the PDF comes from Word

The name in Author usually comes from the source file, so clean that too or every new export will carry it again. In Word for Windows, select File > Info > Check for Issues > Inspect Document, run the inspection, select Remove All next to Document Properties and Personal Information, then export the PDF again. Microsoft’s guide, Remove hidden data and personal information by inspecting documents, has the details. The exported PDF can still record the software that made it and the dates, so check it as above before publishing.

If you use Acrobat

As of September 2026, Adobe’s own guide, Remove metadata from a PDF in 4 steps, says to open Document Properties from the Menu (Windows) or File (macOS), edit or delete the properties, check the fields under Additional Metadata, then save. Adobe’s redaction tutorial, last updated 28 June 2026, also covers sanitising a document to remove “information that is not visible in the file, such as comments, metadata, or hidden layers”, and a note at the top of the tutorial says “Available only in Acrobat Pro, Premium, or Studio.” Whichever tool you use, check the result with something that reads both the document properties and XMP, such as reopening it in Edit PDF metadata or running exiftool -a -G1.

Limits

Metadata cleaning rewrites the descriptive fields. It does not alter the pixels of embedded images — a photograph’s own embedded camera data, if the source application preserved it inside the image stream, is a separate matter — and it cannot change the fonts, layout and phrasing that make a document recognisable. Clean the labels, then read the document as a stranger would.