How to Redact a Screenshot So the Text Can't Be Recovered
Blurring out a card number or pixelating a name feels safe. It often isn't. The original information is still encoded in those blurred pixels, and recovering it can be surprisingly easy. Here's what actually works.
Why blur and pixelation can be reversed
Blur and pixelation aren't deletion — they're math. A blur averages each pixel with its neighbours; pixelation averages blocks of pixels into one colour. In both cases the original values still shape the result. That means the transformation can, in many cases, be partially run backwards.
This isn't theoretical. Security researchers have reconstructed pixelated text, and there are published tools that "unredact" blurred or pixelated regions. It's easiest when the hidden text is short and predictable — a credit card number, an account ID, a name — because there are only so many possibilities to test against the blurred pattern. Exactly the things people most often try to hide.
The only reliable method: a solid fill on a flattened image
Safe redaction has two requirements:
- Replace the pixels, don't transform them. Cover the sensitive area with a solid, opaque block. There's no original data left to reverse because the original pixels are simply gone.
- Flatten the result. Export a single, merged image. If the redaction is a separate layer or object, it can be moved or deleted to reveal what's underneath.
Solid black (or any opaque colour) over the area, exported as a flat PNG, is the standard. Done that way, the covered information is unrecoverable — there's nothing behind the block.
The "box on top" trap
The most common redaction mistake has nothing to do with blur. It's drawing a black rectangle over text in a PDF, Word doc, or slide and sending the file. In those formats the box is a separate object floating above the text — the words underneath are still right there in the file. Anyone can select and copy the text, or just delete the box. Government and legal redaction failures almost always come from this exact mistake. To redact in a document properly, you either flatten it to an image first or use a true redaction feature that deletes the underlying content.
How to redact safely with ClientSide
Go to clientside.sh/redact and drop or paste your screenshot (it takes clipboard paste, so you can grab a screenshot and paste straight in). Draw over each sensitive region. The tool offers three modes — solid black, pixelate, and blur — and for anything you genuinely need hidden, use solid black. The other modes are there for cosmetic cases where reversibility doesn't matter.
When you download, the image is flattened into a single PNG with the blocked pixels permanently replaced — there's no layer to peel back and no original underneath. And like the rest of ClientSide, the image is processed entirely in your browser, so the screenshot you're redacting is never uploaded to a server in the first place.
One more step: strip the metadata
Redacting the pixels doesn't always clear a file's metadata. If your screenshot or image carries EXIF data, run it through clientside.sh/metadata afterwards so you're not leaking through a channel you forgot about. See what your photos secretly reveal for what that can include.
Frequently asked questions
Can blurred or pixelated text be recovered?
Often, yes. Blur and pixelation are reversible transformations, not deletions, and tools exist that reconstruct them — especially for short, predictable strings like card numbers or names. For anything sensitive, use a solid fill.
What is the safest way to redact a screenshot?
Cover the area with a solid, opaque block and export a flattened image. The covered pixels are replaced entirely and merged into one layer, so there's nothing left to recover.
Why isn't a black box in a PDF or slide enough?
The box is a separate object on top of the content; the text underneath is still in the file and can be copied or revealed by deleting the box. Flatten the file to an image or use a true redaction feature instead.
Does redacting also remove the file's metadata?
Not necessarily. Strip metadata separately with clientside.sh/metadata after redacting if the file may contain EXIF data.