PDF Redactor
Draw rectangles to redact sensitive content. Redacted pages are flattened to images so the text cannot be recovered. 100% in your browser.
PDF Redactor
How to redact a PDF
Add your PDF
Drag your file onto the dropzone or click to choose. The tool renders a large thumbnail of every page so you can draw redactions accurately.
Draw redaction rectangles
Click and drag on any page thumbnail to mark a rectangle. The rectangle appears as a semitransparent red box while you're drawing, then as a solid black overlay once you release. You can draw multiple rectangles per page.
Pick a flattening DPI
150 DPI is recommended — sharp text, reasonable file size. 100 DPI for smaller output (text may look slightly soft). 200-300 DPI for print-quality output (larger files).
Apply and verify
Click Apply redactions and download. The output saves as {original}-redacted.pdf. Open the output and verify before sharing. Redacted pages are flattened images — try copying text from a redacted area to confirm it returns nothing.
PDF redaction — done right, not just covered up
The most common — and most dangerous — mistake in PDF redaction is to draw a black rectangle on top of the text without removing the underlying content. The page LOOKS redacted in a PDF reader, but the text is still in the file. Copy + paste reveals it. Text search finds it. Any forensic tool extracts it. It has happened to careful people. In January 2014 The New York Times published an NSA slide deck whose blacked-out agent name and target could be recovered by copying the text, and in January 2019 Paul Manafort's lawyers filed a court document whose "redacted" passages about his contacts with Konstantin Kilimnik came out the same way. Cosmetic redaction is not redaction.
What "destructive redaction" actually means
True redaction means the redacted content is NOT PRESENT in the output file at all. This tool achieves that by re-rendering each redacted page as a flattened raster image (a JPEG embedded in a new PDF page), with the redaction rectangles drawn as solid black on the image. After flattening, there's no underlying text object remaining. Copy + paste from a redacted region returns nothing. Search finds nothing. There's no recoverable content because there's no content there to recover.
If you can copy-paste from a "redacted" PDF and get the original text, it wasn't redacted. It was decorated. Real redaction removes the bytes.
The trade-off: redacted pages lose searchability
Destructive redaction comes with a cost: the redacted page becomes a single image. The remaining un-redacted text on that page also becomes part of the image — no longer selectable, no longer searchable, slightly fuzzier than the original. That's the necessary price of certainty. Pages you don't mark are copied across with their text, fonts and links intact. The output is a new document, though, so what lives at document level does not come with it: bookmarks and the title, author and subject metadata are dropped, and form fields keep showing their values but can no longer be filled in. Only redacted pages are flattened. You get the protection where you need it and the searchability where you don't.
The APAC compliance context
In Singapore, the Personal Data Protection Act 2012 does not mention redaction by name, but two of its obligations bear directly on it. Section 24 requires an organisation to make reasonable security arrangements to protect the personal data it holds, and a PDF whose "redacted" text can still be copied out is not protected. If such a file reaches the wrong people, it is a data breach, and under sections 26B and 26D a breach that is likely to cause significant harm, or that is of significant scale, must be reported to the PDPC no later than 3 calendar days after the organisation assesses it as notifiable. Other jurisdictions have their own rules. Whichever applies, a page that only looks redacted can still leak the data underneath.
What this tool does — and what it doesn't
This tool destructively redacts rectangular regions you draw on each page. The output is a hybrid PDF: pages with redactions are flattened images with black overlays; pages without are preserved verbatim. What it does NOT do: smart text-region detection (you draw the rectangles manually), search for words to redact, or partial-word redaction within a text line (you can draw a tight rectangle that visually covers part of a word, but the line itself becomes a flattened image either way). Annotations and form fields on a redacted page are flattened into its image along with everything else, so cover them with a rectangle if they should go. For non-rectangular or word-aware redaction, Adobe Acrobat Pro's redaction tool is the right choice.
Why a black box on a PDF is not a redaction
Drawing a black rectangle on a PDF page does NOT remove the underlying text. PDFs are layered — text and graphics can sit on top of each other independently.
In January 2019 a court filing by Paul Manafort's defence team was published with black bars over the sensitive passages. The text was still underneath: copying the bars and pasting them elsewhere revealed what his lawyers had meant to withhold.
True redaction removes the underlying bytes. This tool achieves that by flattening redacted pages to images — the text is no longer present in the output.
Flattening means the redacted page becomes a single JPEG image embedded in the new PDF. Text on that page can no longer be selected, searched, or copied — by design.
Non-redacted pages keep their text, fonts and links; form fields on them still show their values but stop being fillable, because the output is a new document. Flattening applies only to pages where you draw rectangles.
This tool flattens at 150 DPI unless you choose otherwise, a default that keeps ordinary body text legible at a moderate file size. Choose 200–300 DPI if the output will be printed or read zoomed in, or 100 DPI for a smaller file.
Adobe Acrobat Pro's Redact tool works differently: it deletes the text and images under each mark and puts a coloured box in their place, so the rest of the page stays searchable text instead of becoming an image. It also removes only visible content, so Adobe tells users to sanitise hidden data such as metadata as a separate step. Our tool flattens the whole page instead, and you draw each rectangle yourself.
Under Singapore's PDPA, a data breach assessed as notifiable must be reported to the PDPC no later than 3 calendar days after that assessment (section 26D). A leaked "redacted" document is assessed like any other breach.
The output is a new document: the original's title, author, subject and keywords are not carried over. File names and anything visible on a page you did not redact are still yours to check.
Always verify the output. Open the redacted PDF in any reader, try selecting text in a redacted region — if you get nothing, the redaction worked. Pasting the redacted region's "text" into a notepad is the simplest forensic check.
FAQ
-
No. pdf.js renders thumbnails, pdf-lib applies the redaction and re-saves — all in your browser. Open DevTools → Network and watch when you click Apply: there's no outbound traffic. For redaction workflows specifically, this matters because the document you're redacting is often the most sensitive one you have.
-
Yes — for the content under the rectangles you drew. Redacted pages are flattened to JPEG images and re-embedded. There is no underlying text object remaining, so copy + paste, search, and forensic extraction all return nothing. Verify with your own copy + paste test on the output before sharing.
-
The whole page is flattened to an image, including non-redacted text. The non-redacted text remains visible (rendered into the image) but is no longer selectable or searchable. This is the trade-off of destructive redaction. If preserving searchability of un-redacted text is critical, accept the cosmetic-redaction risk consciously — but understand it's not real redaction.
-
Yes. Only pages where you draw redaction rectangles are flattened. Pages with no rectangles are copied byte-for-byte from the source — their text remains selectable, searchable, and copyable.
-
150 DPI is this tool's default and suits most documents: body text stays legible and the file stays a moderate size. Use 100 DPI for a smaller file if the text is large. Use 200–300 DPI if the output will be printed or read zoomed in. Higher DPI always means a larger file, and the extra sharpness mostly shows when you zoom in or print.
-
No — metadata (author, creation date, producer, etc.) is preserved. For complete sanitisation including metadata, run our PDF Metadata Cleaner tool on the redacted output. Many genuine document leaks happen via metadata even when the visual redaction is correct.
-
Yes, with caveats. The redaction will visually cover the regions you mark, and the output flattens the image — so the redacted areas are unrecoverable visually. However, if there's no text layer in the source (no OCR was applied), the original was never copy-paste extractable anyway.
-
Open the output PDF in Adobe Acrobat or any PDF reader. Click and drag to try selecting text in a redacted area. If you can select nothing, the redaction worked. Then try a text search for a phrase that should have been redacted — it should not be found. Both tests need to pass.
-
No — not even one that opens without a password. "Restriction-only" protection (an owner password that blocks printing or editing) still encrypts the file's text and images, and pdf-lib, the library this tool uses, cannot decrypt them, so anything it saved would be broken. Since 30 September 2026 the tool says so as soon as you add such a file, instead of building a redacted copy whose untouched pages would not open. Save an unprotected copy first, in software that has the password.
-
Yes on iOS Safari and Chrome on Android. Drawing rectangles by touch works the same as desktop. The thumbnail grid reflows to one column on phones. For accuracy when drawing small redactions, desktop or tablet is recommended.
Method & sources
How it computes
Pages with rectangles are rendered by pdf.js at the chosen DPI (72-300, default 150), the rectangles painted black on that image, and the page replaced by the JPEG; unmarked pages are copied as they are. Export waits for each render to finish and refuses to save if one fails or passes 30 seconds, rather than flattening a half-drawn page.
What this tool implements
- Until this review the export used a ~1 s preview timeout: a slow page could be flattened half-drawn, and a failed render became a blank white page, saved as the redacted file. Tested 2026-09-29: redacted text is absent from the output and unmarked pages keep theirs.
- The output is a new document: title, author, subject and bookmarks are not carried over, and form fields on unmarked pages keep their values but are no longer fillable — the page had said metadata was preserved and form fields kept.
- PDPC's breach guide: a notifiable breach (likely significant harm, or 500 or more individuals) must be reported within 3 calendar days of assessment (section 26D).
Sources
- Personal Data Protection Commission Singapore. Guide on Managing and Notifying Data Breaches under the PDPA (15 March 2021). https://www.pdpc.gov.sg/-/media/files/pdpc/pdf-files/other-guides…
- Mozilla. PDF.js — a general-purpose, web standards-based platform for parsing and rendering PDFs (Apache 2.0). https://mozilla.github.io/pdf.js/
- Hopding A. PDF-LIB — create and modify PDF documents in any JavaScript environment. https://pdf-lib.js.org/
Pick up where you left off
Stored only in this browser — never sent to our servers.