Remove Metadata from PDFRemove Metadata from PDF

Five places where a PDF writes about itself

Properties in a reader show a handful of lines, which makes it look harmless. In reality the file keeps notes in at least five separate places, and clearing one of them does not clear the others.

1. The Info dictionary. The classic list: Title, Author, Subject, Keywords, Creator, Producer and two dates. A document typed by a Mara Okafor in Word will say so here.
2. An XMP packet. A second copy of nearly all of that, written as XML, extended with document and instance IDs, a save history and sometimes rights text.
3. The trailer ID. A pair of numbers. The first one is inherited by every later version of the file, which lets anyone line up copies that were edited separately.
4. Old revisions. Editors often append a change to the end instead of rewriting. What the change replaced is not deleted; it is just no longer displayed.
5. Everything pasted inside. Photos keep their own EXIF and GPS, comments keep reviewer names and times, and design tools keep private blocks with project and version details.

Why strangers' names end up in your document

Nobody types most of these fields. The software that touches the file writes them. Word copies your account name into Author. A scanner announces its model and firmware as Creator. Printing a web page to PDF from a Chromium browser fills Producer with “Skia/PDF” and takes the page title as Title, while the macOS print dialog signs the file with “Quartz PDFContext”. Design programs leave project names behind. The result is a document that names people, tools and companies you never meant to mention.

Deleted text that is not deleted

Take a PDF in which one paragraph was removed and the file saved. The reader shows the new page. The old paragraph, together with the author and date of the previous save, can still be inside the file, because the new version was stacked on top of the old one. Journalists and lawyers have recovered passages this way from documents that were supposed to be final.

Here the file is built again from nothing, carrying over only what the live document needs. In our tests a PDF that had been saved twice came out with only its current text: the earlier wording, the author and the dates were gone.

Why the quick fixes disappoint

• Blanking Author in the properties dialog changes one of two stores. The XMP copy and any old revision still say who you are.
• “Print to PDF” produces a new file with new metadata of its own, and often flattens links and bookmarks.
• Compressing shrinks the file and keeps every property.
• Renaming the file changes nothing inside it.

What works is to rewrite the file and drop whatever the current document does not point to. That is what the button does.

Who benefits most

Scans and photographed pages

A scanner file lists the machine in Creator and Producer and stores every page as a picture. Phone scanner apps do the same and may embed the original photo with its EXIF. The cleaner removes the document fields and the metadata inside those page pictures, but leaves the pictures and any text layer from OCR as they are. Words that are printed on the scanned sheet are content, so a name printed on the sheet stays.

What changes, what does not

Deleted

• All Info dictionary entries, including custom ones such as Company, and every date.
• XMP on the document, the pages and the images, plus private application data.
• The trailer ID.
• Objects nobody refers to any more: superseded revisions, old Info dictionaries, orphaned XMP.
• EXIF, GPS and comments inside embedded JPEGs, with the pixel data copied untouched.

Unchanged

• Pages, text, fonts, images and vector art.
• Links, bookmarks, form fields and the values in them.
• Accessibility tags and the document language (both survived our tests).

Beyond what a metadata tool can do

• Visible content: OCR text, names in headers and footers.
• Author names and dates on comments and annotations, and attached files.
• JavaScript and metadata inside JPEG 2000 images.
• Encrypted PDFs. Remove the password first if you have the right to.

How the rewrite works

Internally a PDF is a heap of numbered objects (pages, fonts, images, streams) that hang off a catalogue, and a trailer points at the catalogue. The tool reads the file, deletes the descriptive entries, walks the document starting at the catalogue and keeps what it can reach. Those survivors are written out as a new file.

Content streams, fonts and images are copied, not rendered and not recompressed. Layouts therefore match, and text stays selectable and searchable. Embedded JPEGs lose their metadata segments with the same lossless method used in the image tool. Since the file is written afresh, its size and checksum change even if little was removed.

Clean a PDF

1. Drop the file or files into the box at the top of the page.
2. Start the cleaning and read the report: every field found, with author, producer and dates marked, plus hidden items such as old revisions and metadata in pasted photos.
3. Download the result or a .zip. It carries the verified clean label only when a second scan of the output finds none of them.

A report, line by line

A typical export from a word processor looks like this before and after (invented values).

• Author: Mara Okafor → removed
• Creator and Producer: Microsoft Word for Microsoft 365 → removed
• CreationDate 2026-09-30 09:14 and ModDate 2026-10-01 17:40 → removed
• XMP packet: 3,412 bytes → removed
• Document ID: one unique identifier → removed
• Earlier revisions: 14 unused objects → removed

Below the list the tool states how many fields are left. Zero means the second scan came back empty.

Trust, then verify

• In a reader: File, Properties shows title, author and dates; some readers show XMP too.
• In a text editor: open the clean file and search for /Author and xmpmeta. We do not hide objects in compressed streams, so a plain search is a fair test.
• On the command line: ExifTool prints everything it finds.

exiftool -a -G1 -s clean.pdf

Your document does not leave the tab

The open-source pdf-lib library, running in your tab, opens the PDF, rewrites it and hands it back as a download. No part of it is sent to our servers, as the network panel of your browser will confirm. That matters most for contracts, medical letters, HR papers and unreleased reports. File size limits: guests 25 MB, free accounts 25 MB, Starter 100 MB, Pro 500 MB, Business 1,000 MB.

Stop leaks at the source

• Microsoft Word: File, Info, Check for Issues, Inspect Document clears personal properties. When saving as PDF, untick “Document properties” in Options.
• LibreOffice: Tools, Options, Security, Options, and tick “Remove personal information on saving”.
• Acrobat Pro: Sanitize Document clears hidden information in one pass.
• Anything else: export, then clean the result here.

Dozens of PDFs

Guests clean 50 PDFs or images a day, one at a time. Free accounts clean 500 a day, 5 at once; Starter, Pro and Business allow 25, 100 and 500 at once with no daily limit. For a whole folder, the API returns a clean PDF and a report per call, and the MCP server lets an assistant such as Claude or Cursor clean PDFs on your computer when you ask. The article on cleaning many files at once compares the options. Photos you share go through the image tool, and the overview of AI metadata covers generation records and Content Credentials.

Questions