Set PDF Metadata So Files Get Found

How to Change PDF Properties: Title, Author, Keywords

Document properties are the only part of a PDF visible before it is opened. The file explorer shows them, the mail client shows them in preview, the operating system's search bar reads them, and the browser tab prints one of them. Which is why a file titled "Contract_final_edits3" with an author of "Ivan" leaves the building more often than anyone would like.

Four fields, and one of them is not what it looks like

The interface offers four text fields: title, author, subject and keywords. The title is not the file name. It is a separate string inside the document, and it is what the browser tab shows when a PDF is opened from a link. The file can be called `scan_0012.pdf` while the tab reads "Reconciliation statement 2026".

Keywords go in as one comma separated line and work for search inside a file store, not on the internet. For a shared drive holding hundreds of contracts, that is a cheap way to make documents findable.

The subject field describes what the document is about in one phrase. Almost nobody fills it in, which is a shame: in mail clients and document management systems it is usually the field that ends up in the description column beside the file name.

There is a fifth field, the creating application, that the tool can also write, but the interface has no box for it: it is reachable only when the service is called programmatically. Worth noticing for a different reason, though, since it is the field that reveals which program the document was built in.

An empty field erases nothing

This is the main trap, and it is worth testing on your own file.

By default new values are merged into the existing ones. A field left empty does not clear the property, it leaves whatever was already in the document. Leaving the author line blank and expecting the name to disappear is the most common mistake made while trying to anonymise a file before sending it.

Tested on a document with its properties filled in: the title was replaced with a new one and the author field was submitted empty, and the name `Ivan Petrov` stayed exactly where it was, along with the subject, the keywords, the creation date and the name of the program the file was made in.

There are two ways to actually make a property disappear. Either type a new value over the old one, or turn on the switch that clears existing metadata.

What clearing the existing metadata really does

The switch is off by default and works more bluntly than one might assume: it does not clear the fields before writing, it throws out the whole property block and then writes back only what you typed.

Zeroing one field alone is not possible: clearing applies to every property at once. If the author has to go while the creation date stays, the date has to be typed back by hand, and the interface has no box for it.

One more detail worth knowing in advance. A PDF records two programs: the one the document was made in, and the one that wrote the file itself. The first, in the table above, is cleared. The second does not stay empty: after processing it carries the name of the library the file was assembled with. The document never becomes fully anonymous, but its link to your working software is broken.

How to set the document properties

1. Open set-pdf-metadata and upload the file. 2. Decide in advance whether you are adding to the properties or replacing them wholesale. 3. Fill in the fields you need. For a full replacement, turn on the clearing switch and type in every value that should survive. 4. Run the job and download the result. 5. Open the downloaded file's properties in your file explorer and check them. That is the only reliable verification.

The form does not show what is already in the file

The fields open empty regardless of what the uploaded document contains. The tool does not read the current properties and does not offer them for editing.

The practical consequence is that the original values have to be looked at before processing, by you. In Windows Explorer that is the file's properties on the Details tab; on macOS it is the Get Info panel. Otherwise you are editing blind, at risk of overwriting something needed or leaving something behind.

The hidden layer that disappears either way

Beyond the familiar property block, a PDF can carry a second one in XMP format. Publishing and office suites write it, and it usually duplicates the title and the author. The problem is that programs show sometimes one block and sometimes the other, so a document that looks scrubbed can go on carrying the old author name in a hidden copy.

Tested: after processing, that block is not in the file at all, with clearing on or off. The document is rebuilt from scratch and the hidden copy of the properties never reaches the result.

For anonymising, that is good news. But if the XMP in your document carried something needed, colour profile data for a printer for instance, losing it comes as an unpleasant surprise.

What survives the operation and what does not

The pages are not redrawn and the content is untouched. The document outline is carried across whole, and fillable form fields, if there were any, keep working: changing a title must not cost a document its structure.

Each value is capped at a thousand characters, and control characters are stripped out of the text. A line break cannot be typed into a title, and that is protection rather than pedantry.

A document under an open password cannot be processed; the job stops and asks for the password. A password that only restricts printing or copying does not interfere with writing properties.

The file itself is reassembled in the process: the pages are moved into a new document one by one. That produces no noticeable change in weight, but the result will not be byte identical to the source even if you changed no field at all.

Properties, privacy and the order of operations

Changing properties is not anonymising. The author's name and the organisation stay in the text of the document, in figure captions, in running headers and in comments. If the task is to remove data rather than rename a file, the work belongs in the content, through redact-pdf, with the properties cleaned as the last step.

Order matters here for several reasons at once.

Properties should be set after the document is assembled. When files are combined with merge-pdf, the result inherits the properties of the first file in the list, not the largest and not the last. Carefully filled properties on every other part are lost.

Signing touches properties too, but carefully. sign-pdf puts the signer's name into the author field only when that field was empty, and appends the signature note to the subject after a semicolon. An author you filled in survives the signature, while the subject comes out longer than you set it, which is worth checking afterwards.

And finally: properties are not protected. Anyone who can open the file can change them. If the content has to stay as it is, that is protect-pdf with permission restrictions, not a carefully typed title.