Redacting PDFs: why black boxes are not enough and how to do it properly
The word "redact" is used loosely, and the difference between its two meanings matters enormously. Covering a line of text with a black rectangle is something anyone can do in a few seconds. Removing that text from the file so it cannot ever be recovered is a different operation entirely.
This guide explains what real redaction does, why cover-ups fail, and how to verify that a redacted file is genuinely clean before you share it.
Redaction vs covering
| Action | Visible result | Underlying text | Recoverable? |
|---|---|---|---|
| Cover with a box | Hidden | Still in the file | Yes — selectable |
| True redaction | Hidden | Deleted from file | No |
| White-out shape | Hidden | Still in the file | Yes |
How true redaction works
Redaction finds the text runs that match the words you target, removes their characters from the page content, and draws a solid black block over the area. The removed characters are gone from the file — copying the page returns nothing where the redaction sits.
This is why redaction is permanent. Once the text is deleted from the output file, it cannot be un-deleted. Keep the original file separately if you think you may need the content later.
What redaction is for
- Personal data such as names, addresses and phone numbers in documents you share.
- Financial figures in contract excerpts or audit reports.
- Identifiers like account numbers and reference codes.
- Any content you are legally required to remove before disclosure.
Word-based redaction
The most useful mode is word-based: list the sensitive words, and every occurrence on every page is redacted. This catches repeated identifiers reliably, unlike manually drawing boxes over each one. Case matching and exact word boundaries matter — check the results to make sure partial matches were handled the way you expect.
Checking the redacted file
- Open the output and visually scan every page for the target words.
- Select-all and copy the whole page; paste into a text editor.
- Search the pasted text for the sensitive words — they must be absent.
- Confirm the black blocks sit exactly over the removed text.
- Verify no other content was accidentally redacted.
Redaction vs other tools
Watermarks label a document but never remove content. Password protection restricts access but a holder with the password sees everything. Only redaction deletes content permanently. If your goal is "these words must never be readable again", redaction is the only honest tool — and it is irreversible, so keep the original elsewhere.
Common redaction mistakes
- Covering with a black box instead of redacting — the text stays in the file.
- Redacting a short form of a word and leaving longer forms visible elsewhere.
- Checking only the visible page and never testing whether the text copies.
- Redacting from a copy and then sharing the original by mistake.
The copy-and-paste test catches the most dangerous failure: a document that looks redacted but still hands out the hidden text. Make it a routine step before sharing anything sensitive, and double-check that the file you are about to send is the redacted version, not the source.
Redaction in legal and compliance workflows
In legal and compliance settings, redaction is often a formal requirement: certain data must be removed before a document is disclosed, and the removal must be verifiable. A word-based redaction pass followed by a copy-paste check is the practical minimum. Keep a record of which words were redacted and on which pages, and retain the original separately where policy requires it.
If a document was only covered rather than redacted, the disclosure may be considered incomplete or defective. That is why the distinction this guide draws is not academic — it is the difference between a compliant disclosure and one that leaks the very data it was meant to protect.
What redaction does not remove
Redaction removes the targeted text from the page content, but it does not touch everything else that might identify the document. File metadata such as the author name and creation timestamps remains unless you clean it separately. Images that contain the sensitive text are not affected by word-based redaction, because the tool works on the text layer, not on pixels. For documents where a name appears inside a photograph or logo, removing that content needs image editing rather than redaction.
- Metadata (author, software, dates) — clean it separately.
- Text inside images and scans — not reachable by word redaction.
- Watermarks and page stamps that could identify the sender — remove them separately.
Privacy
Redaction runs locally in your browser. The document is processed and the redacted copy is generated on your device. Nothing is uploaded, which matters when the document is the very thing you are trying to protect.