How to Redact Sensitive Information From a PDF Properly
Drawing a black box over text isn't redaction - the text is often still there underneath. Here's how proper redaction actually works.
Every so often a redacted government document, legal filing, or corporate report makes headlines for the wrong reason: someone highlighted the black boxes, hit copy, and pasted out exactly the text that was supposed to be hidden. It happens because the person doing the redacting drew a black rectangle over the sensitive text and assumed that was the end of it - visually, it was gone; underneath, it never left the file.
Why a black box isn't redaction
A PDF page is made of layered objects - text, images, shapes - stacked on top of each other. Drawing a black rectangle just adds one more shape to that stack, positioned above the text you want to hide. The text object itself doesn't go anywhere; it's simply obscured by whatever's drawn on top of it. Anyone who selects that area of text, copies the page's text layer, or removes the rectangle object in an editor can get the original content back, because it was never actually deleted - only covered.
This is different from how it looks on a printed, physically blacked-out page, where the ink genuinely obscures the paper underneath. A digital "black box" is more like a sticky note than ink - convincing at a glance, trivial to peel back.
What real redaction actually does
Proper redaction removes the underlying text and image data entirely, not just its visual representation. Instead of drawing something on top, a real redaction tool identifies the exact text or image content in the selected area and deletes it from the document's internal structure, then (often) fills that space with a solid black region purely as a visual marker that content was removed - the difference is that this time there's genuinely nothing left underneath it to recover.
How to redact a PDF step by step
- Identify every occurrence of the sensitive information first - names, ID numbers, account numbers, addresses - since the same detail sometimes appears more than once across a long document.
- Use a dedicated redaction tool (not a highlighter or shape tool) and mark each area that needs to be removed.
- Apply the redaction, which permanently deletes the underlying content rather than just covering it.
- Check the document's metadata (author, company, subject fields) and strip anything sensitive there too - metadata is easy to forget and just as leakable as body text.
- Verify the result by trying to select and copy text from the redacted areas - nothing should be selectable, and searching the document for the removed content should return zero results.
- Save the redacted version as a new file, keeping the unredacted original in a secure, separate location rather than overwriting it.
What kinds of information typically need redacting
- Social security numbers, national ID numbers, and passport numbers
- Bank account and routing numbers
- Home addresses and personal phone numbers in documents going to a wider audience
- Signatures, when a document is being shared more broadly than the original signing purpose intended
- Names of minors or protected parties in legal or medical documents
- Proprietary business information in documents released for a FOIA request or public filing
Don't forget images
Redaction usually gets thought of as a text problem, but sensitive information hides in images too - a photographed ID card embedded in a PDF, a screenshot containing an account number, a scanned check attached to a report. A black rectangle drawn over part of an image has exactly the same problem as one drawn over text: the original pixels are still there underneath, and cropping or covering isn't the same as actually removing that portion of the image data.
Metadata: the part people forget
Every PDF carries metadata alongside its visible content - author name, the software used to create it, sometimes a company name, and occasionally an edit history. For a document being redacted specifically because it's sensitive, that metadata deserves the same scrutiny as the body text. It's common for a carefully redacted document to still list the original author's full name in its properties, which can undercut the whole point of the redaction if anonymity was part of the goal.
Redacting scanned documents
Scanned pages are technically images, not text, which changes the redaction process slightly - there's no underlying text object to delete, since the "text" is really just pixels shaped like letters. Redacting a scan means permanently altering the image data in the redacted region (not layering a shape on top of it), so the original pixels in that area are actually gone, not just hidden behind something opaque.
Redaction tools versus manual workarounds
It's tempting to reach for whatever's already open - a shape tool in a PDF viewer, a highlighter set to black, an image editor - and call it redaction. The problem isn't that these tools can't visually hide text; it's that none of them are built with "permanently delete the underlying content" as their actual function, so it's easy to end up with something that looks redacted but isn't. A tool purpose-built for redaction treats removal as the explicit operation, not a side effect of drawing a shape, which is the difference that actually matters here.
Redacting the same document for different audiences
A single source document sometimes needs multiple redacted versions - a contract shared internally with salary figures visible but sent externally with them removed, or a report released to one regulator with certain sections intact and to the public with those same sections redacted. In cases like this, keep the fully unredacted original in one clearly labeled, access-controlled location, and generate each redacted variant as its own separate file rather than repeatedly redacting and un-redacting the same document - that kind of back-and-forth editing history is exactly the situation where a partially-redacted draft accidentally goes out the door instead of the finished version.
A simple verification habit
Before sending any redacted document anywhere, do two quick checks: try to select and copy text from each redacted area (nothing should highlight or paste), and search the whole document for a word or number you know was in the redacted section (the search should come back empty). These two checks take under a minute and catch the overwhelming majority of redaction mistakes before they leave your hands.
Frequently asked questions
Is drawing a black rectangle over text in a PDF the same as redacting it?
No. A black rectangle is a visual overlay sitting on top of the original text - the text itself is still in the file underneath and can often be recovered by selecting and copying it, or by removing the overlay object in a PDF editor. True redaction permanently deletes the underlying text, not just what's visible.
Can redacted text be recovered from a PDF?
If it was properly redacted - meaning the underlying text and any hidden layers were actually removed, not just covered - no. If it was only visually covered with a black box, yes, often trivially.
Does redaction also need to remove metadata?
For genuinely sensitive documents, yes. A PDF's metadata (author name, company, sometimes editing history) and any embedded images can contain information just as sensitive as the visible text, and a thorough redaction process checks both, not just the main body text.
How can I check whether my redaction actually worked?
Try selecting text in the redacted area and copying it elsewhere - if anything pastes, the underlying text wasn't removed. For a more thorough check, search the document for a word you know was in the redacted section.