Why a black box over a PDF usually hides nothing

A black rectangle drawn over text in a PDF is decoration. The text underneath is untouched, still stored in the file, and usually recoverable by selecting the area and pressing copy. The page looks redacted to a human and is completely legible to a computer.

A PDF page is not a picture. It is a list of instructions describing how to draw the page: set this font, move to this position, draw these characters, draw this line. A reader follows the list from top to bottom and the page appears.

When you draw a rectangle over a word, you do not edit the instruction that drew the word. You add a new instruction to the end of the list: *fill this area with black*. Both instructions are now in the file. The reader draws the text, then draws the rectangle on top of it, and you see a black box.

Anything that reads the file rather than looking at it — a copy-and-paste, a search index, a script, a screen reader — walks the same list and finds the text sitting there in plain view. The rectangle is just another shape; it does not mean "ignore what is beneath me", because PDF has no way to express that.

This is not a beginner mistake. Court filings in the Paul Manafort case were published with passages blacked out, and readers recovered them by copying the text out of the PDF. A batch of released Justice Department documents was un-redacted the same way by members of the public within hours.

In each case the people doing the redacting had looked at the result and seen black boxes. That is the trap: the only feedback the tool gives you is visual, and the failure is invisible to the eye by definition.

1. 1

   Try to select the hidden text

   Open the PDF, click just before the black box, and drag across it as though selecting a sentence. If a selection highlight appears over or past the box, there is text there. Copy it and paste it somewhere to see what it says.

2. 2

   Search for it

   Use the reader’s find function and search for something you expected to remove — a surname, a number. A hit inside a redacted area means the text is still in the file.

3. 3

   Extract everything

   The two checks above depend on your reader. To see everything a machine can pull out of the file regardless, open the redaction tool and press “Show me what a machine can read”. It lists every string in the document, page by page. Nothing is uploaded.

Real redaction means the instructions that drew the text are no longer in the file. There are two honest ways to get there.

Whichever route a tool takes, the part that matters is what happens next: it should read the finished file back and confirm there is nothing left. A tool that says “redacted” without looking is giving you the same assurance as the black box did.

The redaction tool on this site flattens the pages you mark, leaves the pages you do not alone, and then re-reads the result and shows you the count of readable items on each redacted page before you download anything.

A scan is a photograph of a page, so there is often no text layer to leak and a black box genuinely does cover everything.

The exception is a scan that has been through OCR, which many scanners and document systems apply automatically. OCR adds an invisible text layer over the image so the document can be searched — and that invisible layer behaves exactly like ordinary text. If you can search a scanned document, it has a text layer, and a black box will not hide it.

• Metadata. Title, author, the software that made the file and sometimes the original filename travel with the document regardless of what is on the page. A file named settlement-draft-confidential.pdf has already said something.
• Other copies of the same fact. Removing a name from page 3 does nothing about the same name on page 11, in a header, or in an embedded attachment. Redaction is per-occurrence, not per-fact.