Redacting personal data (PII) from a PDF or image, explained

PII — personally identifiable information — is anything that points to a specific person: a name, a home address, an e-mail address or phone number, a national ID or passport number, a date of birth, a bank account. Before you share a contract, a medical letter, a log file or a screenshot, those details often have to go. The Eraser finds them and covers them with black boxes, entirely inside your browser.

How it finds personal data

For a PDF with real text, it reads the text layer directly, which is exact. For a scan or an image it runs OCR (Tesseract.js) in your browser first; pick the document’s language for the best result. The text is then checked in three ways:

• Checksums. An IBAN must pass the mod-97 test and have the right length for its country; card numbers must pass the Luhn test; a Belgian national register number (rijksregisternummer, also the BIS variant) must pass its mod-97 check; a Dutch BSN must pass the 11-test; US social security and UK national insurance numbers must fall in valid ranges. Random numbers that merely look similar are left alone.
• Context. Dates are only hidden when a birth label is next to them (“date of birth”, “born”, “geboren op”, “né le”, “Geburtsdatum”), so invoice and meeting dates stay. Passport, ID card, driving licence, patient, customer, employee and policy numbers are hidden when their label is present.
• Names are found after a title (Mr, Mrs, Dr, Dhr., Mevr., M., Mme, Herr, Frau…), after a label (“Name:”, “Naam:”, “Nom :”, “Patient:”), after a greeting (“Dear”, “Beste”, “Bonjour”) or above a sign-off line, and when a common first name is followed by a surname. Once a surname is known, it is blacked out everywhere else in the document too.

Why you should still review

No automatic redaction is perfect. A name without any of those clues — a colleague mentioned in the middle of a sentence — can be missed, and OCR can misread a scan. That is why the result is a review list: untick anything that is not personal data, type an extra word or phrase to black it out everywhere, or drag a rectangle on the page to cover a signature, a photo or anything else by hand. Click a box on the page to switch it off again.

Safe output

The downloaded PDF or PNG is built from images of the redacted pages. The original text layer is not copied into it, so the hidden text cannot be selected, searched or copied back out — unlike drawing a black rectangle in a PDF editor, which often leaves the text underneath.

Privacy

Your file never leaves your device: PDF.js and Tesseract.js run in the browser. The only network requests load those libraries and the OCR language data. Nothing is stored; closing the tab clears everything. This is a practical tool, not a certified compliance product — for regulated documents, have the result checked as your rules require.

More free tools