Redact a PDF

Yes, on every page you put a box on. Those pages are rebuilt as pictures, the black is painted into the pixels, and the original text is not written into the new file at all. After building the file, the tool reads the redacted pages back and saves it only if they hold no text. With the searchable setting on, it saves it only if no text sits over a box and no word found only under a box appears anywhere in the file. Pages without a box keep their text, so check that you boxed every occurrence.

Yes. Type the name, number, or phrase and press Box every match, and every copy on every page gets a box, ignoring capitals. Each box can be moved, resized, or removed like one you drew. A scanned page holds a picture of its words, so search finds them there only with Also search scanned pages ticked, which reads those pages with OCR first.

Yes. Tick Match as a pattern under the search box, type the shape of what to find, and press Box every match. A pattern is a short description of text rather than the text itself. \d stands for any digit, a number in curly brackets repeats what comes before it, and + means one or more. So ACC-\d{6} finds ACC-482913 and every other ACC- with six digits after it, \d{3}-\d{2}-\d{4} finds anything shaped like 123-45-6789, and [A-Z]{2}\d+ finds two letters followed by digits. Capitals are ignored, as in the plain search, and a hyphen in the pattern matches the dashes PDFs print too. Patterns are JavaScript regular expressions, and the regex tester is a good place to try one on sample text first. A pattern that could match nothing at all, such as \d*, is refused, because it would box the gaps between letters, and so is one that matches more than 2,000 places in the file or runs for more than 3 seconds. A pattern reads each page as one line of text, with line breaks as spaces, so it can match across a line break but not from one page to the next.

Yes, with Also search scanned pages ticked. A page that stores no text of its own is drawn as a picture and read with OCR, and both the text search and Find personal details then look through the words it read and box them like any other match. The Scanned pages figure beside the page shows how many pages have no text, and the setting greys out when there are none. The engine is about 4.5 MB, downloaded the first time you use it, and each page takes a few seconds, but a page is only read once while the file stays open, so a second search is quick. OCR can misread or miss a word, so the note after a search says how many boxes came from recognised text, and those pages are worth checking by eye. A page with typed text and a scanned picture on it counts as text, so words in the picture still need a box drawn by hand.

Yes, in one press. Under Find personal details, tick the kinds you want and press Box them all, and every email address, phone number, card number, IBAN, or US Social Security number of those kinds gets a box, on every page. The page then says how many of each were found. They are ordinary boxes, so you can move, resize, or delete any of them, and one Undo takes the whole search back. Card numbers must pass the check digit every card carries, and IBANs their own check digits, so long invoice numbers are not taken for either. A phone number has no check digit, so it is found when it is written the way numbers are dialled: in groups, such as 020 7946 0958 or (555) 123-4567, or after a country code such as +44. Dates, prices, postcodes, and invoice or order numbers are left alone. A number written as one long run of digits with no + in front, or a local number of fewer than eight digits, is not found; type it into the search box instead. Like search, it reads a scanned page only with Also search scanned pages ticked.

200 dpi is sharp enough to read and print ordinary documents and keeps the file a reasonable size. Choose 300 dpi for small print, fine drawings, or anything that will be printed professionally, and 150 dpi when the file has to be small, such as for an email with a size limit. The resolution only affects the pages you redact.

A page stored as text and fonts is compact. The same page stored as a picture holds every pixel: at 200 dpi, a few tens of kilobytes for a page that is mostly white and a few hundred for one full of print or pictures. The file grows by that much for each page you redact. Choosing a lower resolution makes it smaller.

Open this page in any current browser, on Windows, a Mac, Linux, a Chromebook, or a phone, and drop the PDF in. Draw boxes over what has to go, or search for a name and box every copy, then press Redact and download. There is nothing to install and no account, and the file stays in the tab. The result is a normal PDF that any viewer opens.

Yes. A box covers any area of the page, whatever is under it: a photo, a signature, a stamp, a column of figures, or a block of text. The black is painted into the picture of the page, so what was there is not in the new file. A signature or a stamp has no words for search to find, so drawing by hand is how those are covered.

Yes. The new file is built from scratch, so it starts without the old document properties, such as the title, author, and subject, and without bookmarks, form fields, and attachments. Those can repeat the text you are hiding. Pages without a box are copied across as they were, so anything else on them stays.

Yes. A scan is already a picture of each page, so nothing is lost by redacting it, and the boxes are painted into the new picture the same way. Some scans carry an invisible text layer added by text recognition, and that layer is removed from redacted pages too. Tick Keep redacted pages searchable to have a new one made from the redacted picture. To find a name on the scan before redacting, tick Also search scanned pages, and the search reads the pages with OCR.

It reads each redacted page as a picture, after the black is painted in, and lays the words it recognises over the page as invisible text. Find, copy, and screen readers then work on those pages again, and the page looks exactly as before. It cannot bring back anything under a box: the recognition sees only the finished picture, where that part is solid black, and the tool checks the file for hidden words before saving it. It can misread words, most often in small print, unusual fonts, or a poor scan, so text you copy or search for may differ from what the page shows. A word right against a box is the likeliest to lose a letter or go missing, because it runs into the black. It reads English, and pages drawn at 300 dpi read best. The engine is about 4.5 MB, downloaded the first time you use the setting, and each page takes a few seconds. It is the same one the PDF OCR tool uses.

Yes, in the phone’s own browser on an iPhone or an Android phone, with nothing to install. Hold a finger still on the page for a moment, then drag to draw a box, and drag the box or one of its corners to adjust it. A finger that moves straight away scrolls the page as usual. On a small screen, search is the easier way in: type the name or number, press Box every match, and every copy gets a box without any fiddly dragging.

Not while the password is on. Remove it with the PDF password remover first, which needs the password, then redact the unlocked copy. Protect the redacted file again afterwards if it needs a password. A file that opens freely but is locked against printing or copying is refused as well, and the remover lifts that only with the file’s owner password.

No, the page takes one PDF at a time. Redact it, download the result, then drop in the next file. The search box, the pattern, and the personal details you ticked stay as they are, so the same search can be run on each file. To join the redacted files afterwards, use the PDF merger.

No. The pages are drawn, blacked out, and put back together in this browser tab. For a document that needs redacting, sending it to a server first would defeat the purpose.