Redact PDF

Most ways of blacking out a PDF only put a shape on top. A black rectangle from a drawing tool, a highlight set to black, or a white box from a whiteout feature hide the text on screen, but the text is still in the file underneath. Anyone can select it, copy it, search for it, or open the file in another program and delete the box. Our own Add Text to PDF whiteout works that way too, which is why it carries a warning.

Real redaction removes the content. This tool does that for every page you put a box on.

1. Each page with at least one box is drawn to an image by the same PDF engine your browser uses to show it, at the resolution you chose. Form fields and comments on that page are drawn too.
2. The boxes are painted into that image in solid black or white. Pixels under a box are replaced, not layered.
3. The page is rebuilt from scratch: its old text, drawings, fonts, images, links and comments are removed and replaced by the one image.
4. All pages are copied into a brand-new PDF. Only what the remaining pages actually use is written out, so the old content of redacted pages does not survive as unused leftovers inside the file.
5. Images that a redacted page used and that other pages merely listed without drawing are removed from those other pages as well.
6. The new file is opened again and its text is extracted to check the result.

We tested this with a three-page PDF containing a name, an account number, a code word and a photo. After redaction, extracting text from the redacted pages returned nothing, and a scan of every decompressed stream in the file found none of the removed words. The third, untouched page still had its normal selectable text.

Turning a page into an image is what makes the removal certain, and it has costs you should know before sending the file:

• Text on redacted pages can no longer be selected, copied or searched, including the parts you did not cover. Screen readers cannot read those pages.
• Those pages are a little less sharp when zoomed in far, and the file usually gets bigger. In our test a lightly filled A4 page at 200 dpi in JPEG came to about 90 KB; pages full of text or photos are larger.
• Links on redacted pages stop working.

Pages you did not touch keep everything: selectable text, vector sharpness and links. If only one page of a 40-page contract needs redaction, only that page changes. If someone needs searchable text later, they can run PDF to Text with OCR on the result, which will read everything still visible and nothing under the boxes.

Search looks at the PDF's own text layer and boxes every exact match, on every page. Spaces in your search are flexible, because PDFs often store words without the spaces between them, so John Smith also matches JohnSmith. Matching ignores upper and lower case unless you switch on Match case. Whole word stops Ann from matching inside Annual.

It cannot find:

• Text in scanned pages or photos of documents. These have no text layer; the tool tells you which pages have no text so you can box them by hand.
• Different spellings, abbreviations or formats. Searching 010-1234-5678 will not find 01012345678. Search each form you expect.
• Text drawn as shapes, which some design programs produce.

Box positions for search matches are calculated from the font metrics in the file and padded slightly, so they usually cover the whole word with a small margin. Look at each one: in an unusual font a box can come out a little short, and you can drag its handle to extend it.

After the file is made, the tool reopens it and reports three things. First, how many characters of text can still be extracted from the redacted pages; this should be 0. Second, whether any word you searched for still appears anywhere in the file. Third, whether text you covered by hand also appears on a page you did not redact, for example a name on page 1 that is repeated in a footer on page 6. If that warning appears, search for the text, mark the other pages, and create the PDF again.

The check only reads text. It cannot tell whether a box fully covers a signature or a photo, so open the downloaded file and look at every redacted page before you share it.

JPEG keeps the file smaller and suits pages with photos. PNG keeps text edges perfectly crisp and suits plain text pages, but pages with photos become large. Very large pages, such as posters, are reduced automatically so a single page never exceeds about 36 megapixels, which keeps phones from running out of memory.

Because the result is a fresh PDF, some document-level extras are left behind on purpose: bookmarks, document properties such as title and author, embedded attachments, and fillable form behavior. Form fields on untouched pages keep their appearance but stop being editable. If the original had a password, the redacted copy does not; add one with Protect PDF if needed. Links to web addresses on untouched pages keep working, and so do most links that jump between pages.

For photos and screenshots rather than PDFs, blur or pixelate areas with Blur Image. To remove whole pages instead of parts of them, use Delete PDF Pages, which does not turn anything into images.