Permanently redact a PDF.
"I drew a black rectangle over the sensitive text in Preview" is the most-viraled data leak of the last decade. The text is still selectable, still copy-pasteable, still searchable. Real redaction removes the underlying text — usually by rasterizing.
The problem with rectangles
Drawing a black shape over text in a PDF viewer (Preview, Acrobat annotations, Word "black highlight," anything similar) adds a black object on top of the text. The text underneath is unchanged. Any of these recovers it:
- Select-all + copy → paste into a note app.
- Open the PDF in a code editor as raw bytes — the text is right there as UTF-8 or embedded font glyphs.
- Use a "remove annotations" tool — the black shape is an annotation.
- OCR the page — even if the text object were removed, OCR would re-read the pixels.
This is how Manafort, the NYPD, the CIA, and countless law firms have leaked redacted content. The mistake is universal.
Real redaction: rasterize
/pdf-redact/ does the correct thing:
- Draw the redaction on each page. Mouse or touch. The tool overlays a transparent black preview.
- On export, each page is rasterized — rendered to a bitmap at high resolution — then the black boxes are drawn as opaque pixels. The final PDF contains only images, no text objects.
- Verify. After download, open the redacted PDF in any viewer and try Select-All + Copy. The blacked-out region returns nothing. OCR on the black region returns nothing. There are no annotations to strip.
Step-by-step
- Open /pdf-redact/. Drop the PDF.
- Navigate pages with Prev / Next.
- Draw a rectangle over every sensitive region. Names, SSNs, account numbers, addresses, signatures.
- Undo if you slip. Clear-page to restart a page.
- Export. Rasterization + composition happens on export — a "before" preview intentionally shows the transparent overlay so you can double-check placement.
- Verify the export. Open the downloaded PDF, try Select-All, try OCR on a redacted region. Should be blank.
What about metadata?
Rasterizing removes the PDF's text-layer but not its top-level metadata (Author, Creator, ModDate). Chain into /pdf/'s "Strip metadata" mode to clean those too.
Legal / evidentiary use
Rasterized redaction is what the U.S. National Archives' technical guidelines recommend for FOIA releases. It's what courts accept as "true redaction." It's not, however, cryptographically signed — for court-of-record use, you may still need a signed hash of the redacted PDF or a chain-of-custody workflow that browser-based tools can't provide.