Stop using the digital Sharpie.

That black box you draw over sensitive text in most PDF tools? The text is still there, selectable and copyable. It's how the TSA leaked an airport screening manual, and how a 2019 court filing exposed the exact data a defense team tried to keep sealed. A black rectangle isn't redaction. This tool actually removes the content, then checks its own work before handing back the file.

Text removed from the file, not covered
Hidden OCR text layers on scanned pages checked too
Metadata, annotations, and hidden layers stripped
No trace left in the file's revision history
Can't verify a page is clean? It refuses the file, doesn't guess
Processed and discarded — nothing retained

Redact a PDF

Upload a PDF, mark the regions to remove, download the result. Nothing is kept.

This actually removes the underlying text and image data, not just a visual black box. After processing, this tool re-checks its own output and refuses to return a file it can't verify is clean. See the FAQ below for what's in scope for v1.
or drag and drop a PDF
No file selected

FAQ

Is this really free?

Yes, completely, no limit.

Is my data stored?

No. Your file is uploaded, processed, and discarded right after your download completes, nothing is retained.

How is this different from other free redaction tools?

Most "redaction" tools draw a black box over text but leave the underlying content extractable via copy-paste or search, a documented failure that has burned journalists, the TSA, and court filings alike. This tool actually removes the text and image content in the marked region, then re-checks its own output and refuses to return a file it can't verify is clean, rather than risk shipping incomplete redaction.

What's out of scope in v1?

Pages rotated 90 or 270 degrees aren't supported yet, rotate them to standard orientation first (see PDF Tools). PDF form field values and complex vector graphics aren't scrubbed. Redacting a tagged (accessible) PDF removes its accessibility structure tree entirely, since that structure can carry duplicate text outside the normal content stream. If a page can't be verified clean, it's rejected instead of guessed at.

Any limits?

25MB per file, up to 20 pages.