A black box is not a redaction
Drawing a black box over text is an expensive mistake made with PDFs, and it is made constantly by people who are being careful.
Draw a black rectangle over a name in almost any PDF editor and you have added a rectangle. The name is still there, in the file, underneath it. It is still selectable, it still turns up in a search, and dragging across the black box and pressing copy pastes it out in full. The document looks perfectly redacted on screen, which is the whole problem.
It is not a theoretical risk. It is how the unredacted parts of the Manafort court filing were read within minutes of publication, how confidential figures came out of the Facebook and BlackBerry filings, and how a steady stream of council FOI responses and NHS trust reports have leaked names that somebody thought they had removed. Every one of those looked black.
How this tool works
The page is redrawn as a picture with the black boxes painted into it before it is written back out. There is then no text left on that page to leak, because there is no text on that page at all. It is the same result as printing the page, blacking it out with a marker and scanning it back in, without the printer or the marker.
Only the pages you have boxed are redrawn. Everything else in the document is copied across untouched, so a sixty page bundle with two redactions keeps fifty eight pages searchable and the file stays roughly the size it was. Rasterising the whole document would be simpler and would leave you with a worse document.
Then it checks, which is the part nobody does
When it has finished, the tool reopens the file it just made, reads the redacted pages back and counts the characters of text still on them. That number is on screen. It should be zero, and you are looking at a measurement rather than a promise.
It also searches the rest of the document for every word it removed. This catches the mistake that actually leaks redactions: the name is blacked out on page three and printed in full in the distribution list on page nine. That is not a fault in the redaction, it is a second copy, and nothing else will tell you it is there.
Search finds it, including the pieces
Type a name, an address, a phone number or a price and every occurrence in the document is boxed at once. Doing that by eye across a long document is where redactions get missed.
It searches the joined-up text rather than the raw pieces, which matters more than it sounds. A PDF frequently stores a word split across several items, so "Fletcher" can be held as "Flet" and "cher". A search that looks at each piece finds nothing, and the user concludes their own name is not in their own document.
What it does not do
It cannot find what it cannot read. On a scan there is no text at all, so search does nothing and you box things by eye. It says so rather than reporting no matches, which would be true and misleading at the same time.
It does not know what is sensitive. Nothing here decides for you which name matters, and there is no automatic detection of personal data, because a tool that half finds those is worse than one that leaves the judgement with you.
Metadata is cleared as part of the job, so the author name, the title and the timestamps go with it. That has caught people out on its own: the document was clean and the properties still said who wrote it and when.
Redaction is the one where uploading defeats the point
The document is opened, redrawn and checked on this page using your own machine, which you can confirm in your browser network tab. Nothing is uploaded and nothing is stored anywhere.
That matters more here than on any other tool on this site. A document you are redacting is by definition one containing something that must not get out, and the usual way to redact one for free is to upload it to a company that keeps it for a few hours and asks for your email.
Common questions
How do I properly redact a PDF for free?
Drop the PDF in, drag a box over anything that has to go or search for it by name, then press redact. The page is redrawn as a picture with the boxes painted into it, so the text underneath is destroyed rather than covered. Nothing is uploaded and there is no account.
Why is drawing a black rectangle over text not enough?
Because the rectangle is a separate object sitting on top. The text is still in the file underneath it, still selectable, still found by a search, and it comes straight back out with copy and paste. It looks completely black on screen, and that is why the mistake keeps being made.
Can the text be recovered from a file redacted here?
Not from the redacted pages. They are no longer text at all, they are pictures, so there is nothing left to recover. The tool proves it by reopening the finished file and reading those pages back in front of you.
Does it work on a scan with no text in it?
Yes. Drag boxes over what needs to go. Searching will not work because there is no text to search, and the tool says so rather than quietly finding nothing.
What does redacting cost me?
Only the pages you box are redrawn, and those pages stop being selectable text. Everything else in the document is left exactly as it was, so it stays searchable and the file stays small. Metadata is cleared at the same time.