Find text hidden in a PDF

The redaction that is not a redaction

The most common way confidential text leaks from a PDF is also the simplest: someone draws a filled rectangle over a name and saves the file. The rectangle is a drawing. The name is still in the page, underneath it, and anyone can select straight through the box and copy it. Governments and law firms have released documents this way repeatedly. This tool looks specifically for text that sits beneath something painted after it.

Why white-on-white happens by accident

It is rarely malicious. Template text gets recoloured instead of deleted, a light theme is applied over dark text, or content is pasted from a slide with a dark background. The result is the same either way — invisible on screen, fully searchable in the file. We report white text even when there is a filled shape behind it, because that shape might be white too, and say which case we found.

What it does not do

It reports; it does not remove. Deleting hidden text means destroying part of the page, which is what redaction does, and that is deliberately a separate tool with its own confirmation. Keeping the two apart also means you can check whether this is telling the truth about a document you know, before you trust it with one you do not.

Not a malware scanner

This finds concealed text. It makes no judgement about intent and it does not look for exploits. Text hidden in a document is often entirely innocent — a printer mark, a template leftover, an accessibility label. What it gives you is the ability to look before you send.

Questions

What counts as hidden?

Five things: text painted white on an unfilled page, text with zero opacity, text positioned outside the page boundary, text drawn before an image that covers it, and text below about two points — too small for any reader to resolve at any zoom.

Will it flag a caption written over a photograph?

No. The check is paint order, not overlap: text drawn after an image sits on top of it and is visible, while the same text drawn before it is buried. Only the second is reported. If we cannot determine the order, we report it and say so, because a missed finding is worse than an explained one.

Does it find hidden images or hidden layers?

Not yet — this release covers text. Attachments, embedded scripts and document properties are reported by the Inspect tool, which looks at the file rather than the pages.

How many pages does it scan?

The first two hundred. Beyond that the panel tells you the analysis was partial rather than implying the rest is clean.

Is the document uploaded to scan it?

No. It is read in your browser. A document you suspect of carrying something is precisely the document you should not be handing to a third party, which is why this tool exists here rather than on a server.