Redact PDF
Find phone numbers, IDs and other sensitive data and black them out. The text is truly deleted, not covered by a box. Files never uploaded.
Click to select or drop a PDF here
Detection and redaction run entirely in your browser — never uploaded
Not sure nothing is uploaded? Take the airplane-mode test
This page is a PWA — once it has loaded, its assets live on your device. Turn on airplane mode or unplug the network right now, come back, and drop a file in: detection, redaction and export all still work. That is a check you can run yourself, without trusting a promise or reading a network panel.
- 1Load this page normally and let it finish
- 2Switch on airplane mode / disconnect Wi-Fi
- 3Drop in a PDF, redact and download as usual
One exception: OCR for scanned pages downloads a ~8MB model on first use, and works offline from then on.
Three common situations
HR: background checks, payslips
Background reports and payslips usually carry ID numbers, bank cards and phone numbers, while the recipient only needs the name and dates. Auto-detection surfaces those three; confirm them and redact in one pass, and draw boxes for the free-form figures a payslip adds.
Legal: case files and exhibits
Material filed with the other side or the court often needs the identities of parties and third persons removed, and it is usually scanned. Scanned pages are located with local OCR — note that a redacted page becomes an image, so say so if the recipient requires searchable text.
Finance: statements and receipts
When a statement goes to auditors or a counterparty, account numbers go and amounts stay. Card numbers are verified with Luhn, so invoice numbers are not mistaken for cards, and digits in adjacent table cells are not merged into a false match.
Why drawing a black box is not redaction
Drawing a black rectangle over sensitive text feels like redaction, and it is the most common and most dangerous mistake in this task. The box is just a drawing instruction layered on top: the covered text objects remain untouched in the content stream, so a text extractor, a plain copy-paste, or simply dragging the box away brings the original back. Real redaction has to remove those characters from the file. This tool renders the page to a bitmap, paints over it, and rebuilds the page from that bitmap — the redacted text physically does not exist in the exported file, regardless of which reader opens it.
Redacted text is genuinely gone
Areas are burned into pixels and the page is rebuilt, so even pdftotext extracts nothing. The trade-off is that the redacted page becomes an image.
Auto-detect, then fill the gaps yourself
Phone numbers, Chinese ID numbers (checksum verified), bank cards (Luhn verified) and emails are found automatically; drag on the page to mark anything else.
Scans work too
Pages with no text layer are read with local OCR to locate text — the model and the recognition both run inside your browser.
Files are never uploaded
Detection and redaction happen entirely in your browser. Documents that need redacting are exactly the ones that should never be uploaded.