Why a redacted PDF so often still contains the text

The usual way to redact a PDF is to draw a black rectangle over the passage, in a PDF reader or an image editor, and save the file. The rectangle is a drawing added on top of the page. The words stay in the file underneath it, and a copy-paste, a text extractor or simply opening the document in another program brings them back. Nothing about the appearance of the file gives this away: it looks exactly like a document that was properly redacted.

This page reads the file and tells you what is still there. It is the same detection the redaction tool uses — the same rules, the same checksums — pointed at a document it did not produce.

What it checks

What it cannot tell you

It cannot tell you a document is safe to send. Finding nothing means the rules found nothing — they validate checksums, and most of what makes a document sensitive has no checksum at all. It also reads only the file’s text: a passage that was flattened into an image is not extractable, so it will not appear here, and looking at the page is the only way to know whether it is still legible to the eye.

That limit is the same one the redaction tool states about its own work, and for the same reason. The guide on badly redacted documents sets out what stays reachable in a file and what to check before sending one.

Is the file uploaded?

No. The document is read by this page, in your browser. There is no account, no server and no storage — you can load the page, disconnect from the network, and check the file offline. No file is produced either: this tool reads and reports, and writes nothing. How to tell whether a tool really is local explains how to verify that for this page or any other.

If it found something

The document has to be redacted again, properly — covering the passage a second time changes nothing. Caviardeur removes the text from the file rather than hiding it: the page carrying a redaction is turned into an image, the zones are painted onto that image, and the image replaces the page, so there is no text layer left to recover anything from. It is free, it needs no account, and the file never leaves your browser there either.