Clean a PDF before you send it

A PDF carries more than its pages. The author’s name and the software that made it. An XMP packet repeating those in a second place. Scripts. Attachments. Comments somebody left in review. Links to places you did not intend to send anyone. And, if the file has been saved a few times, the earlier versions of itself, still in the file, still readable.

Most of that is invisible in a reader and all of it travels when you send the document.

Read, remove, read again

The reading is the part that makes this worth using rather than trusting.

The document is inspected first, so there is a record of what it was carrying. The removable things are taken out. Then the finished file is inspected again by the same reader, and the result reports both halves: what was removed, and what is still there.

That second list is the honest one. A tool that says “cleaned” and hands back a file still carrying an open action has told you something false about a document you are about to send, and there is no way to notice from the outside. Reading it back is how the claim becomes checkable.

What goes

The document information fields and the XMP metadata packet. Both, because they hold the same author name in two places and clearing one is exactly how a name survives a cleaning somebody believed in.

Document JavaScript, and any action set to run when the file opens.

Embedded attachments, which a PDF can carry invisibly and which most readers will happily extract.

Annotations: comments, sticky notes, stamps, highlights and link targets. If a reviewer left a note in the margin, it goes.

Form fields are flattened rather than deleted, so the values stay on the page and stop being editable. A filled application form keeps its answers.

And earlier revisions. A PDF saved repeatedly can append rather than rewrite, keeping every prior version inside the same file. Rewriting it drops them, which happens here as a consequence of saving properly rather than as a separate step.

What stays, and why

Two of the things the inspection can find are not touched.

Content outside the crop box. A crop hides without deleting, so material can sit just off the visible page. Removing it means re-cutting the page, and that is a decision about what the document should say.

Text hidden underneath something. A black rectangle over a name is not a redaction; the characters are still there. Fixing it properly means removing those characters and proving they are gone, which is what Redact a PDF does. Doing it silently here would be the worse of the two options: it looks like a cleaning tool and behaves like a redaction one, and nobody would know which they got.

Both are named in the result rather than left out of it.

When to use it

Before sending a document outside the organisation that made it. Before publishing one. Before filing one anywhere the file itself, rather than a printout, is the thing being submitted.

If you only want to see what is in a file without changing it, Inspect a PDF reads and reports and writes nothing.

Nothing leaves the device

The document is read, rewritten and read back by a worker inside your browser. No upload, no account, no server copy, which matters more here than almost anywhere, because the whole reason to run this is that the file contains something you would rather not hand over.