How to Remove Metadata From a PDF (See What Yours Leaks First)

Quick answer: Every PDF carries a hidden set of document properties: typically a Title, Author, Subject, Keywords, the software that created it, and two timestamps. Your name is very often in the Author field without you ever typing it there, because Word and most office apps fill it from your account. You can check it in seconds (File > Properties in Acrobat, Tools > Show Inspector in macOS Preview), and you can remove it either by clearing the fields in an editor or by rebuilding the file so the old properties never come along. Below is what a PDF really stores, shown on a file we built for this article, and the removal routes that hold up.

Photos leak location and camera details through EXIF. PDFs have their own version of the same problem, and it gets far less attention, which is exactly why it bites people: a resume that names a different author than the applicant, a "fresh" report whose creation date is three years old, a confidential draft whose title still says draft.

What a PDF actually stores about you

PDF metadata lives in two places. The one everybody hits is the Info dictionary, a small set of key-value fields defined by the PDF standard. The second is an XMP packet, an XML block Adobe introduced that can mirror and extend the same fields. If XMP sounds familiar, it is the same standard photos use alongside EXIF; we cover that family in EXIF vs IPTC vs XMP.

To make this concrete, we generated a PDF the way an office app would, then read its properties back. This is everything the file volunteered:

Field	What our test file carried
Title	Q3 Financial Review CONFIDENTIAL
Author	Jane Smith
Subject	Internal audit draft
Keywords	confidential, draft, audit
Creator	Microsoft Word for Mac 16.88
Producer	macOS Quartz PDFContext
Created	2026-03-14 09:30
Modified	2026-08-01 14:00

Read that table the way a stranger would. The Author is a full name. The Title says confidential. The Creator names the exact software version, which dates your toolchain. And the two timestamps tell a story on their own: this "new" document was created in March and quietly edited in August.

None of this shows on any page of the PDF. It only shows in the properties panel, which means the sender almost never sees it and the recipient absolutely can.

First: look before you send

Checking costs nothing and settles whether you have a problem at all.

• Adobe Acrobat or Acrobat Reader: File > Properties. The Description tab lists Title, Author, Subject, Keywords, the application, and both dates.
• macOS Preview: Tools > Show Inspector. The first panel shows the document's properties.
• Most other readers (Firefox, Edge, and the rest) expose the same fields under a Document Properties or Document Info menu item.

If the Author field is empty and the dates do not embarrass you, you may not need to do anything. Empty is a perfectly common state; plenty of generators write nothing.

Method 1: clear the fields in an editor

If you have an editor that can write document properties (Acrobat's paid tiers can, as can LibreOffice and many PDF utilities), the direct route is to open the properties dialog, delete the contents of Title, Author, Subject and Keywords, and save a copy.

Two things to know about this route:

• Clearing the visible fields does not always clear the XMP packet. The Info dictionary and the XMP block are separate structures, and a lazy edit can blank one while the other keeps the old values. This is why the verification step at the end matters more than which tool you pick.
• The Producer field often refuses to stay empty, because whatever software saves the file tends to stamp its own name there. A Producer that names a PDF library is harmless; the fields that identify a person are the ones worth fighting for.

Method 2: rebuild the file so nothing carries over

The more reliable route is not to edit the old file but to produce a new one. A rebuilt PDF starts with fresh properties belonging to whatever produced it, and the original Author, Title and dates simply never make the trip.

We verified this behavior directly: taking the test file above and re-saving it through a rebuild pipeline left every identifying field blank. The new file knew nothing about Jane Smith, the audit, or the March creation date.

Ways to rebuild:

• Print to PDF. On both Mac and Windows, printing a PDF into a new PDF creates a file whose properties describe the print pipeline, not the original document. Page content survives; document properties do not carry over.
• A converter that reconstructs the file. Our Compress PDF tool rebuilds the document page by page in your browser, so the output carries none of the input's properties. One honest caveat that matters: it re-renders each page as an image, which means selectable text becomes part of the picture. For a scanned document that is no loss at all. For a text contract you need to keep searchable, use the field-clearing route instead.
• Image round trip. For single-page or image-heavy documents, exporting pages as images and recombining them with Image to PDF produces a file with no ancestry. Same trade-off: the text layer is gone.

The right choice depends on whether the text layer matters. Rebuilding is the cleaner strip; editing is the one that preserves searchability.

What removal does not fix

A few things survive every metadata pass, and pretending otherwise would be doing you a disservice:

• Content is content. A name typed on the page, a header with your company logo, a signature: none of that is metadata, and no metadata tool touches it.
• Redactions have their own rules. Covering text with a black rectangle in a viewer does not delete the text underneath. If you need true redaction, use a tool that actually removes the content, not one that draws over it. For images inside the document, permanent pixel redaction is the standard we apply in our Pixelate and Redact tool: the export rewrites the pixels, not a layer above them.
• Copies you already sent are already sent. Metadata removal is a before-sending habit, not an after-the-fact fix.

PDF vs photo metadata, in one minute

If you already think about EXIF, the mapping is straightforward. The Author field is the PDF's version of the camera owner tag. Creator and Producer are the Make and Model. CreationDate is DateTimeOriginal. The practical difference is volume: a photo can carry GPS coordinates and hundreds of technical tags, while a PDF usually carries under a dozen fields. But the sensitive ones, name and dates, are just as loud. Photos get stripped with an EXIF remover; PDFs need one of the routes above because the container is different, even though the privacy problem is identical.

Frequently asked questions

Does my PDF definitely have my name in it? Not definitely, but check before assuming either way. Office suites fill the Author field from the account name by default, so documents born in Word or similar apps usually carry one. PDFs generated by web services often carry none.

Can people see PDF metadata without special tools? Yes. Every mainstream PDF reader shows document properties through a normal menu item. No forensics involved; it is two clicks.

Does converting to PDF from Word remove the Word metadata? No, usually the opposite: the export writes your account name into the PDF's Author field. The conversion is where the metadata gets in, not where it comes out.

Will removing metadata change how the PDF looks? The field-clearing route changes nothing visible. The rebuild routes preserve the page appearance but flatten selectable text into images, so pick based on whether you need the text layer.

Is PDF metadata ever useful to keep? Inside your own archive, yes: real creation dates and authorship make files findable and provable. The point is control at the boundary. Keep it internally, strip it when the file leaves.

Bottom line

A PDF tells more than its pages do: an author name you did not type, software versions, and honest timestamps that can contradict the story around the file. Check the properties before you send anything that matters, clear the fields when you need the text layer intact, rebuild the file when you want certainty that nothing carries over, and always reopen the result to confirm. The check takes ten seconds and it is the only step that never lies to you.