How to redact a document properly (and how people get it wrong)
Redacting something looks like the simplest task in the world: drop a black rectangle over the name, the address, the account number, and you're done. Except a startling number of redactions fail — the hidden text is still sitting in the file, one copy-paste or one deleted box away from being read. Courts, governments, and major companies have all leaked sensitive information this exact way. Here's why it happens, and how to redact so the hidden part is genuinely, permanently gone.
The core mistake: covering isn't removing
Almost every failed redaction comes from the same misunderstanding. In most document and PDF editors, when you draw a black box, you're adding a new object on top of the page — like laying a sticky note over a word. The word underneath hasn't gone anywhere. It's still in the file's text, fully intact, just visually hidden. Anyone who opens that file can often select the "blacked-out" text and copy it straight out, or simply delete the box and read what's beneath.
The same trap catches images. Paste a black rectangle as a layer in an image editor and save in a format that keeps layers, and the original is preserved underneath. The redaction looks perfect on screen and protects nothing.
The one principle to remember: real redaction removes the underlying content, it doesn't just hide it from view. If the original still exists anywhere in the file, the redaction has failed — no matter how solid the black box looks.
How this goes wrong in real life
This isn't a theoretical risk. There's a long, almost comic history of high-stakes documents released with black boxes over sensitive passages, where anyone could copy the text out from underneath or the covering layer could be peeled away. Legal filings, official reports, and corporate disclosures have all spilled names, figures, and secrets this way — because someone trusted a black rectangle that was only skin-deep. The lesson each time is identical: the visible cover-up and the actual data are two different things, and hiding one doesn't remove the other.
The blur trap
Blurring has its own version of the problem. A gentle blur or a mild pixelation looks like it destroys the content, but the original detail is often still faintly encoded in the softened pixels — and in some cases it can be partially reconstructed. There have been real instances of "blurred" or lightly pixelated text and faces being recovered well enough to read or identify. A blur that's too weak is closer to a hint than a redaction.
The fix is to make sure the detail is truly destroyed, not just softened. A strong mosaic that averages each block down to a single flat colour throws the underlying detail away for good — you can't reconstruct a hundred different pixel values from one average. A solid block is even more absolute, since it replaces the pixels entirely.
The reliable way: flatten it into an image
The most dependable way to redact is to remove the layers problem entirely by working with a flat image. When a redaction is painted directly onto the pixels of an image and the image is then re-saved as a brand-new file, there is no separate layer to peel back and no hidden text to copy — because the covered pixels have been overwritten and the original no longer exists in the output. What you see is genuinely all there is.
This is exactly why turning a sensitive document into an image and redacting that is such a robust approach: images don't carry a hidden text layer, and burning the redaction into the pixels leaves nothing underneath.
Redact an image the safe way
Black out or pixelate any part of an image, and the redaction is burned permanently into the pixels — no layer to remove, nothing recoverable — with the file's metadata stripped too. It all happens in your browser; the image is never uploaded.
Open the blur & redact tool →Don't forget the metadata
One more leak worth closing. Even a perfectly redacted image can betray you through its hidden metadata — most notably the GPS coordinates many photos carry, recording exactly where the picture was taken. It would be a poor result to carefully black out a face while leaving the location of the person's home buried in the file. A good redaction step re-saves the image in a way that clears this metadata along with the visible content, so nothing sensitive rides along invisibly.
A quick checklist
- Never trust a black box alone in a document or PDF editor — assume the text underneath is still there unless the tool explicitly removes it.
- Avoid weak blur or light pixelation for anything sensitive; use a strong mosaic or a solid block.
- Flatten to an image and burn the redaction into the pixels, then re-save as a new file, so there's no layer to recover.
- Strip the metadata so location and other hidden details don't leak.
- Check the result — try to select or copy the redacted area, and confirm nothing comes out.
The bottom line
Redaction fails when people confuse covering information with removing it. A black box on a layer, or a blur that's too gentle, hides the content from a casual glance while leaving it fully recoverable underneath. Do it properly — destroy the underlying detail, flatten it into the pixels, re-save as a new file, and clear the metadata — and the hidden part is truly gone. When something is sensitive enough to redact, it's worth making sure the redaction is more than skin-deep.