What redaction actually removes
The difference between hiding something and removing it, and why so many redactions have failed in public.
Redaction failures are a recurring news story: a court filing, a government release, a corporate disclosure, published with names visible to anyone who copied the text out from under the black boxes. The cause is always the same misunderstanding.
What a page actually is
A PDF page is a list of drawing instructions: place this glyph here, fill this rectangle, draw this image in this box. What you see is the result of running that list.
Adding a black rectangle adds one instruction to the end of the list. It does not remove any of the others. The instruction that places the word is still there, so:
- Selecting across the box and copying gives you the text
- Searching for the word finds it
- Any tool that reads the text rather than rendering the page sees it
- Converting the document to Word or plain text produces the word in full
The box is a curtain, and the curtain is not attached to the window.
What a proper redaction does
It removes the content — the actual instructions that place those characters, or that part of that image — and then puts a mark where they were. What remains is a page that has no knowledge of what used to be there.
| Approach | On screen | Text still in the file? | Safe to disclose? |
|---|---|---|---|
| Draw a black rectangle | Looks redacted | Yes | No |
| Draw a rectangle and flatten | Looks redacted | Depends on the tool and the page | Not something to gamble on |
| Delete the whole page | Page is gone | No, but so is everything else on it | Only if nothing else on the page was needed |
| Crop the region away | Region not shown | Yes — cropping hides | No |
| Redact | Looks redacted | No | Yes, once verified |
Flattening turns pages into pictures, which does remove the underlying text of those pages — but it is a side effect of a feature built for a different purpose, applied page-wide, and it destroys searchability and accessibility as it goes. It is not a controlled removal of specific content, and a confidential disclosure is not the place to rely on a side effect. Use redaction.
The four places content survives a careless redaction
1. The document's own details
Every PDF can carry a title, an author, a subject and keywords. None of it appears on any page. All of it travels with the file. A document titled “Henderson settlement — confidential draft” has already said more than the redacted body. Sanitize clears it.
2. Annotations and comments
A note left during an internal review — “can we settle at 40?” — is stored as an object in the document, not as part of the page. It survives anything done to the page. Sanitize clears comments and marks along with the details.
3. Text inside images
A name in a photograph, a screenshot of an email, a scanned signature block: these are pictures, so searching for the name finds nothing and text removal removes nothing. They have to be found by eye and redacted as image regions.
4. The obvious places nobody looks
Headers, footers and stamps repeat on every page, which is exactly why people stop seeing them. A matter reference in a footer, a client name in a header, a “prepared for” line on page one. Check the first page and the last page specifically, and check one page from the middle.
Verifying a redaction
A redaction is not finished when it is applied. It is finished when it has been checked:
- Search the redacted copy for every term you removed. Nothing should be found.
- Select across the redacted areas and try to copy. Nothing should come out.
- Check the document's own details — title, author, keywords.
- Look at every page, particularly for images and for headers and footers.
- Check the file size. A redacted document that is exactly the same size as the original deserves a second look.
Why local processing matters more here than anywhere else
The document you are redacting is, by definition, the most sensitive document you handle. Uploading it to a web service in order to make it safe to disclose is a contradiction: the unredacted version has now been on somebody else's machine. Verso does redaction on your Mac, and contains no code that talks to the internet at all.
Frequently asked questions
Why is a black box not a redaction?
Because a PDF page is a list of drawing instructions and the box adds one rather than removing any. The text underneath is still in the file and can be selected, copied, searched and converted.
Does flattening remove redacted text?
It turns pages into pictures, which has that effect as a side effect — applied to a whole page, destroying searchability and accessibility. It is not a controlled removal and not something to rely on for a confidential disclosure.
Does cropping remove content?
No. Cropping changes what is shown; the content is still in the file and another tool can reveal it by widening the crop.
What gets missed most often in a redaction?
The document's own title and author, comments left during review, names inside images, and headers and footers that repeat on every page.
How do I verify a redaction?
Search the result for everything you removed, try to select and copy across the marks, check the document's details, and look at every page. If nothing comes back, it worked.
Is it safe to redact a document using an online tool?
It means uploading the unredacted version — the most sensitive copy of the document — to someone else's server. Redacting on your own machine avoids that entirely.