Why the black box in Word and Google Docs does not work
You have a document with something in it that should not go out. You select the text, set the highlight colour to black, and it disappears. Or you draw a filled rectangle over it. The page looks right, you export a PDF, and you send it.
The words are still there. Anyone who receives the file can get them back in about four seconds, and they do not need any tools to do it.
What each of those tools actually does
The black highlighter
Highlighting sets a background colour behind the characters. The characters are unchanged and still in the document; they are simply the same colour as what is behind them, or close enough that you cannot read them. Select across the area, copy, and paste into anything at all, and the text arrives in its new home in whatever colour that program uses.
This is the one that has put real names into court records. Eugene Volokh wrote up a filing
where a lawyer had used the Google Docs black highlighter and the document went out with,
in his words, the underlying text still remained in the PDF document
.
The black rectangle
A filled shape is an object placed on the page at a position, with a fill colour and a stacking order. It sits above the text in that order, so it hides it. It does not touch it. Exporting to PDF preserves both the rectangle and the text underneath, because both are part of the document.
Black text on a black background
The same idea reached from the other direction, with the same result, and a worse failure mode: some viewers and most text extraction ignore colour completely.
Deleting the text and typing blocks over it
This one does remove the words, and it introduces a different problem. Programs that track changes or save by appending can keep the earlier state inside the file, so the version with the words in it travels along underneath the version without them.
Why careful people are caught by this
Because every program any of us has used has taught us that the page is the document. A commenter quoted by Volokh said it better than any security guide has managed:
Users are trained WYSIWYG. What you see is what you get. Quoted by Eugene Volokh, writing about redaction failures in court filings
A word processor is built on that promise and keeps it everywhere except here. There is no warning, the menu item does not say that the text remains, and the result looks identical to the result of doing it properly. The people this happens to are usually being careful. They are doing the thing the interface offered them.
It does not help that the word for the correct operation is one most people never meet. Searching for how to black something out returns highlighting tutorials, because that is what the words mean everywhere else.
How to see it for yourself, on a document you already have
- Take any document where you have blacked something out with a highlighter or a shape, and export it to PDF as you normally would.
- Open the PDF in whatever you usually read PDFs with.
- Select the whole page — drag across it, or press Ctrl+A / Cmd+A — then copy.
- Paste into a plain text editor such as Notepad or TextEdit. Not back into Word, which will paste the formatting and hide it from you again.
- Read what arrives. If the covered words are in there, they were in the file you were about to send.
This takes under a minute and it is worth doing once on a real document, because the surprise is what makes the habit stick.
What to do instead
The words have to come out of the file. There are two ways to be sure of that.
A tool with a genuine redaction function. Some PDF software deletes the underlying content rather than covering it. Check the output rather than the label, because the feature is easy to name and harder to implement, and researchers in 2023 assessed eleven widely used tools and found every one of them leaking in one way or another. The eight things a copy-paste check will never catch are set out separately.
Rebuilding the page as an image with the boxes painted into it. Nothing is left to recover because there is no text left in the file. That is what Redaction Studio does, and the method is explained here, including what it costs you — the output is a picture, so it cannot be searched and screen readers cannot read it.
Whichever you use, check the file afterwards. Covering and removing look identical on screen, which is the whole problem, so the only way to tell them apart is to inspect what is in the file.
Before you send it
Check the document — free, no account, nothing uploaded. It reads the file and tells you whether the words you covered are still in it.