PDF metadata: the author name lives in two places

Guide · 18 August 2026

In one sentence. A PDF stores its metadata twice — once in the /Info dictionary, once in an XMP stream — and most cleaning tools empty only the first, so the author's name quietly stays in the file.

What a PDF knows about you

Open the properties of almost any PDF and you will find a small set of fields that were filled in without anyone asking:

  • Author — a person's name, almost never typed in by hand.
  • Title, subject and keywords — often the working title of an early draft, which is not always the title you published under.
  • Creator — the application the document was written in.
  • Producer — the software that converted it to PDF, usually with its exact version.
  • Creation and modification dates — down to the second.

Where the name comes from

Hardly anyone fills in the Author field; the software does. Export a PDF from Word and the field usually carries the name of the account the application is signed in with. Export from InDesign or another licensed tool and it is typically the name the licence was registered under. That is how an invoice, a report or a CV sent as "final_v2.pdf" ends up naming a person — or the person's employer — that the visible pages never mention.

The first place: the /Info dictionary

The traditional home of PDF metadata is a dictionary called /Info. The entries that matter for privacy are eight: Author, Title, Subject, Keywords, Creator, Producer, CreationDate and ModDate. This is what most "document properties" dialogs display, and it is the part that most cleaning tools edit — because it is the easy part.

The second place: the XMP stream

The same file usually carries a second copy: an XMP stream, a block of XML embedded in the document and referenced by the /Metadata entry of the document catalog. It generally duplicates the author and the software fields, and often adds information of its own. A PDF viewer is free to read either copy — which is exactly why cleaning one and not the other achieves so little.

The classic mistake, in two variants

The first variant is stopping at /Info. The XMP stream lives elsewhere in the file, a tool that never looks at the catalog never finds it, and the author's name survives the cleaning in the second copy.

The second variant is subtler: replacing values with an empty string instead of deleting the entry. An empty string is still a present entry — some readers display the empty row, and the entry's existence is itself information. QuietMeta removes the entries outright rather than blanking them.

What QuietMeta removes

You choose by category, and each category maps to specific entries: author; title, subject and keywords; software, meaning both Creator and Producer; and the two dates. When the selection touches authorship, description or software, the XMP stream is removed as well, and the report says so in as many words — because that block duplicates the author and software fields.

Encrypted PDFs are declined rather than forced: QuietMeta does not guess passwords. The inspection report says the file is password-protected and that its metadata could not be read, and cleaning refuses the file outright rather than pretending it was cleaned.

Nothing is re-encoded

Removing the entries means the file has to be written out again, so the bytes of the cleaned PDF differ from the original. But unlike a recompressed JPEG, a rewritten PDF loses nothing: the text, the fonts and the embedded images are copied as they are. What your reader renders is the same document, minus the fields you asked to remove.

One detail says a lot about how careful a tool has to be here. The library QuietMeta uses would, by default, stamp a fresh modification date on the document the moment it is opened. QuietMeta disables that — in the inspector and in the cleaner — so that reading your file does not silently modify it, and cleaning does not reintroduce the timestamp it just removed.

Proof, not promise

After cleaning, QuietMeta reads the produced file again with the same reader used for inspection, and shows you what is present now: which fields remain, and whether an XMP block is still there. Not a claim that the cleaning worked — a fresh reading of the actual result.

The limits are stated just as plainly. Metadata removal does not touch the visible content of the pages: a name typed into the document, a signature block, an address in a header — all of that stays, because it is the document, not data about it. If the pages themselves name you, cleaning the metadata is only half the job.

Checking yours

Drop a PDF on the inspection page and you will see the author, the software that created and produced the file, the title and keywords, the dates, and whether an XMP metadata block is present — then choose the categories to remove. Reading also works on images and the office formats; Word and OpenDocument files carry author names too, in the document properties of the archive.

Files sent to the web version are held in memory for one request and released — never written to disk, a database or a backup. For whole folders at once, or for documents you would rather not send anywhere at all, the desktop application does the same work offline.

Inspect a PDF Get the desktop app

Related