PDF Redactor — black out text, images and metadata_

A PDF redactor removes sensitive content from a PDF instead of just hiding it. Here you mark what has to go — drag over text, click a word, draw a box over a signature or a photo, search for a name across the whole document, or let the scanner find e-mail addresses, phone numbers, card numbers and IBANs for you.

When you save, every page that contains a redaction is rebuilt as an image with the marked areas painted over, and the rest of the file is written from scratch. The text objects, the pixels and the metadata that were on those pages are not part of the new file at all — there is nothing left underneath to copy, select or recover.

Drawing a black rectangle in Acrobat Reader, Preview or Word only paints over the screen. The characters are still in the file: select the area, copy it, and the text comes back — this is how newspapers and courts have leaked redacted documents for twenty years. True redaction deletes the content itself. That is what happens here, and after saving the tool re-opens the file it just produced and tries to extract the redacted text again, so you can see the proof rather than take it on trust.

No. The PDF is opened with pdf.js and rebuilt with pdf-lib, both running in this browser tab. There is no back end to send a file to. Open your browser's network tab while you work and you will see nothing leave — the only requests are for the page and those two libraries. Close the tab and the document is gone from memory.

Free, with no account, no watermark, no daily quota and no file-size ceiling other than your device's memory. Most online redaction tools either upload your file, cap you at one document a day, or keep the good parts for a paid plan. There is no server bill here, because your machine does the work.

Yes — until you save. Undo and redo (Ctrl/⌘+Z and Shift+Ctrl/⌘+Z) walk back through every mark, a selected box can be deleted with Delete, and the redaction list lets you remove any single one. The saved file is final: keep your original if you might need it.

Three ways, and they can be mixed. Drag across the text in text mode to mark every run you touch. Click a single word to mark just that word. Search for it and press redact all matches to catch every occurrence on every page — that is the safe option for a name that appears forty times in a contract.

Switch to box mode and drag a rectangle over it. Boxes are not limited to text: whatever is under the rectangle — text, photo, signature, barcode, handwriting on a scan — is painted over before the page is turned into an image, so the pixels themselves are replaced.

It reads the text layer of every page and flags patterns: e-mail addresses, phone numbers, credit-card numbers (validated with the Luhn check, so random digits are not flagged), IBANs (validated with the mod-97 check), US social security numbers, Dutch BSN-style national IDs (mod-11 check), EU VAT numbers, dates of birth, IP addresses, URLs and amounts. Each type shows a count and a few examples, and one click marks every occurrence. It is a helper, not an oracle — read the document as well.

Yes. Draw the box once — over a header, a footer, a letterhead or a column of names — and press repeat on all pages. There is also redact whole page for pages that have to disappear completely.

Yes. Type a label such as FOIA (b)(6), GDPR art. 4 or PRIVILEGED and it is drawn inside the box in the saved file. Leave the field empty for a plain black box. White boxes are available for redactions on dark or coloured backgrounds, and a blue box for the "marked for review" convention some firms use.

Yes, with one difference: a scan is a picture, so there is no text layer to search and the personal-data scan has nothing to read. Drawing boxes works exactly the same, and because the page is rebuilt as an image the covered pixels are genuinely gone. If you need to search a scan, run it through OCR first.

Pages with no redactions are copied across as they are, so their text stays selectable. Pages with redactions are re-rendered at the resolution you choose, the boxes are painted on, and the result is placed in the new file as an image. Then the whole document is written fresh: metadata, XMP, attachments, document JavaScript and bookmarks are simply never carried over, and annotations are dropped too unless you untick that option.

If you leave keep the remaining text searchable ticked, yes. After the page image is placed, the text that was not redacted is written back invisibly at its original position, so search, copy and screen readers still work while the redacted words are absent. Untick it for the strictest possible output, where a redacted page carries no text at all. Text that the built-in font cannot represent (for example Chinese, Japanese, Arabic or Cyrillic) is left out of that invisible layer rather than placed wrongly.

The tool checks for you. After saving it re-opens the produced file, extracts all of its text, and searches for every string that was under a redaction. The result is shown as a short report: what was checked, whether anything leaked, whether the metadata is empty and how many pages were rebuilt. You can repeat the check yourself with any PDF reader's search, or with pdftotext.

Usually yes, on the redacted pages: a page of text is a few kilobytes, the same page as an image is a few hundred. Choose 150 dpi for screen use, 200 dpi as a balance, or 300 dpi when the file will be printed, and tick JPEG for photos and scans — it is much smaller than PNG. Pages you did not touch are untouched, so a 200-page report with two redacted pages barely grows.

They are dropped. When the file is loaded the tool reports what it found — document metadata (title, author, producer), an XMP block, embedded attachments, document JavaScript, annotations, form fields, bookmarks — and none of that is carried into the saved file, because the file is rebuilt rather than patched. That matters: hidden metadata has revealed authors and revision histories in plenty of published documents.

Annotations (links, comments, highlights, form fields) are removed by default, because comments and form values are a classic leak. Untick remove annotations and form fields to keep them on pages you did not redact. On a redacted page they are gone either way — the page becomes an image.

A tool cannot make you compliant, but the technique is the one those rules ask for: the information is removed, not covered. Because nothing is uploaded, no third party ever holds the unredacted document, which is usually the harder part of the argument. What remains your responsibility: redacting everything that identifies a person (including in metadata and file names), following your organisation's procedure, and checking the result before you hand it over.

• A black rectangle drawn over text that was never removed — copy and paste reveals it.
• Redacting the body while the metadata, the bookmarks or the file name still carry the name.
• Forgetting that the same detail appears in a header, a footer, an appendix or a second copy inside the file.
• Pixelating or blurring instead of covering — blur can sometimes be reversed.
• Sending the redacted PDF together with the original attachment.

The search, the pattern scan and the post-save check on this page exist because of the first three.

Not from the saved file. The redacted page exists only as an image with the covered pixels painted over, and the original page objects are not in the document. What you must not do is re-use the original file afterwards, or send both.

Yes, once the page is rebuilt: the pixels under the box are replaced, not hidden. Blurring or pixelating is the risky approach — it keeps a degraded version of the data, and pixelated text has been reconstructed before.

Nothing but your theme and language choice. The document, the redactions and the saved output live in memory only, and a reload clears them. There is no history and no local copy of the file.

Standard PDFs from version 1.0 to 2.0, including scans and generated reports. Password-protected files must be unlocked first — a PDF that needs a password to open cannot be read here. Very large documents (hundreds of pages, or tens of megabytes) work, but rendering is done by your device, so be patient on a phone.

One document at a time. For a repeatable job across many files, the same approach — pdf.js to find the text, pdf-lib to rebuild the pages — scripts well in Node.js locally.

B box mode, T text mode, ←/→ page, +/− zoom, Delete removes the selected box, Ctrl/⌘+Z undo, Shift+Ctrl/⌘+Z redo, Ctrl/⌘+F jumps to the search box, Ctrl/⌘+S saves the redacted PDF.

Once the page and its two libraries have loaded, yes — the rest happens locally. Reloading with no connection will fail, because pdf.js and pdf-lib are fetched from a CDN; open the page before you go offline and it keeps working.

English, Dutch, French, German, Spanish, Portuguese, Russian, Japanese, Arabic and Indonesian — pick one from the globe in the top bar, or add ?lang=nl to the address.