How to redact an image safely: blur, pixelation or solid coverage?

To hide a name, address or account number in a screenshot, cover it with an opaque block and export a flattened image. Leave a margin around the text, then open the downloaded file to check the edges and any repeated details.

The image redaction tool starts with Rectangle and Solid block selected. Those are useful defaults for covering text.

Cover the whole detail, including its edges

1. Open or paste the image. Keep the original separately.
2. Choose Solid block and draw a rectangle around the sensitive detail. Include punctuation, letter edges and a little surrounding space.
3. Zoom in. Move the rectangle or use Bigger in the region list if anything extends beyond it.
4. Check the rest of the image for the same information: quoted replies, headers, browser tabs, notifications and background labels.
5. Export as PNG for a screenshot with sharp text, then reopen that download.

The Treatment control applies to newly drawn regions. Changing it to Solid block does not convert an existing blurred region. Delete and redraw that region, and check its treatment in the list.

Rectangles are easier to inspect than a scribbled line over lettering. If you use Freehand, check for gaps between strokes. For several related redactions, similarly sized blocks can reveal less about the lengths of the hidden names than boxes fitted tightly around each word.

Why blur and pixelation are poor choices for private text

Blur blends neighbouring pixels; pixelation reduces a region to larger blocks. Both discard some detail, but their output still depends on the original. A result that you cannot read by eye can retain clues about the letters underneath.

Researchers have recovered blurred and mosaiced text under a range of conditions, including known fonts and layouts. That does not mean every blur can be reversed or that a reconstruction recovers the exact original pixels. The practical concern is that enough information may remain to identify a word or number.

An opaque fill replaces the covered pixels in the flattened output. It does not remove clues elsewhere in the image: an uncovered email domain, face, uniform or conversation can still identify someone. A partly transparent marker leaves a mixture of the original and marker colour, so use a fully opaque cover for details you intend to remove.

A saved image is different from an editable overlay

In an editor, a rectangle may be a separate object sitting above the original. Sending the editable document can let someone hide or remove that object.

ToolsNow paints its regions into a new image when exporting. The saved PNG, JPEG or WebP does not include the editor’s region list or an underlying image layer. Your original file and the working session still exist separately; share the exported copy.

A PDF needs particular care. Drawing a rectangle on a PDF page or changing its crop boundary can leave the text underneath in the file. Use a dedicated PDF redaction feature that removes the selected content, or remove it in the source document and make a new PDF. Redaction does not have to rasterise an entire document when the editor can remove objects properly.

Check what remains outside the covers

• Visible clues: names in other messages, reflections, addresses, notifications, faces and distinctive surroundings.
• The filename: this export keeps a cleaned-up version of the original name and adds -redacted. A person’s name can survive that change.
• Metadata: this tool re-encodes the image rather than copying the source’s EXIF or GPS fields. You can inspect the saved copy with the photo metadata viewer. Encoder-generated fields may still be present.
• Document data: if you are sharing a PDF as well, check its metadata, comments and attachments separately.

Inspect the export at full size, not only the reduced preview. Check both the coverage and the information you meant to leave readable. Neither a successful download nor a metadata check establishes that the whole image is anonymous.

Sources and further reading

• Hill, S., Zhou, Z., Saul, L. & Shacham, H. “On the (In)effectiveness of Mosaicing and Blurring as Tools for Document Redaction.” PoPETs, 2016
• NSA — Hidden Data and Metadata in Adobe PDF Files: Publication Risks and Countermeasures
• UK Information Commissioner’s Office — anonymisation guidance
• NIST SP 800-122 — Guide to Protecting the Confidentiality of Personally Identifiable Information
• CIPA DC-008 — Exif 2.32 specification

Last reviewed 2 October 2026.