Can someone get the text back from under the boxes?

Not from the pages you marked. Those pages are rebuilt from a flattened image with the boxes already painted in, and the original page is never copied into the new file, so there is no hidden layer to recover. The finished file is also searched for the text that was under your boxes, and if any of it is still findable you get an error rather than a file.

Why can I no longer select the text on that page?

Because the page is now a picture. Taking text off a page without rebuilding it is exactly what leaves recoverable copies behind, so losing the text layer on a redacted page is the price of the redaction being real. Pages you did not mark keep theirs.

Why is my file so much bigger?

An image of a page takes far more space than the text on it, and only the pages you marked are affected. If size matters more than crispness, drop the quality setting to 150 DPI, which is enough for reading on a screen.

Does it remove my name from the document properties?

Yes. Title, author, subject, keywords and producer are all cleared in the output, because those fields often carry a name or an internal path that nobody intended to send on.

It refused to give me a file. What went wrong?

The check after the rebuild still found text that your boxes were meant to remove, so nothing was handed over. This is deliberate: a file that looks redacted and is not would be worse than an error message. Make sure the boxes fully cover the text, including any part of a line that pokes out of the edge, and try again.

Can I redact a signature or a photograph?

Yes. The boxes are painted into the pixels of the page, so anything under one goes, whether it was text, a signature, a stamp or part of an image.

Redact PDF

Drag a box over anything you want gone. The pages you mark get redrawn without the covered content in them, so there is nothing left underneath for anyone to copy out, and the finished file is searched before you are given it.

Choose a PDF or drop it hereOne file, up to 50 MB. Nothing is uploaded.

Runs entirely in your browser. Your files are never uploaded.

How to use Redact PDF

1Pick a PDF, or drag it onto the box above.
2Click a page in the strip on the left, then drag across whatever has to go. Marks can be taken off again.
3Set the box colour and, if you want one, a label like REDACTED.
4Press the redact button, read what it says it is about to do, and confirm.

Why a black box on its own hides nothing

Drawing a filled rectangle over a line of text in a PDF editor changes what the page looks like and nothing else. The words are still in the page, sitting under the rectangle in the order they were written. Select across the area and paste it somewhere, run any extraction tool over the file, or open it in software that ignores the drawing order, and the text comes straight back out.

Documents have been published this way by courts, police forces, government departments and newspapers, and the information under the boxes was read within hours of release each time. The people who made those files were not careless with the rectangles. They were working with a tool that only drew them.

That is the implementation this tool refuses to ship. It also avoids the other half-measure, which is deleting the text operators and leaving the rest of the page as it was, because a page stripped that way can still carry the words in a form that gives them back.

What happens to the file

Every page you marked is drawn to pixels at the quality you chose, and your boxes are painted into those pixels. The page is then rebuilt in a new document from that flattened picture, and the original page, with its text and its drawing instructions, is never copied across. There is no layer to switch off and nothing underneath to select, because what you are given was built from an image of the page taken after the boxes went on.

Pages you did not mark are copied over whole. Their text stays selectable and searchable, which is why marking one line on page three of a forty-page contract costs you the text layer of page three alone.

The document properties are cleared on the way through: title, author, subject, keywords, and the name of the software that made the original. Those fields routinely carry a person, an internal file path or the name of a client, and whoever is sanitising a document rarely means to pass them along. Bookmarks, attachments and form fields are left behind as well, because the output is built from scratch rather than edited.

The check that decides whether you get a file at all

Before a page is redrawn, the text sitting under each of your boxes is read and written down. Once the new file has been built, it is opened again and those same pages are searched for every piece of that text.

If any of it can still be found, you get an error saying so and no file is produced. A redaction tool that cannot verify itself is only able to tell you it drew a box, so this one reports what it searched for and what came back instead of asking you to take the result on trust.

The check has one limit worth stating. It can only look for text that a PDF reader can read in the first place, so on a scanned page, where the words are already part of a picture, there is nothing to search for. The summary says so in those cases rather than implying something was proved that was not.

What a rebuilt page costs you

A rebuilt page is a picture of a page. Nobody can select its text, search it or copy a paragraph out of it, and that applies to you as much as to whoever you send it to. Screen readers lose the text on those pages too, which matters if the document has to stay accessible.

The file also gets bigger, often by a lot: a page of plain text might be a few kilobytes as text and several hundred as an image. There are three quality settings for that reason. 150 DPI suits something that only needs reading on screen, 200 is fine for ordinary use, and 300 is there for documents that will be printed or carry small print you cannot afford to soften.

If the rest of the document needs to stay searchable, mark as few pages as the job allows. One box on one page leaves the other thirty-nine as they were.

What it will not do for you

It does not go looking for anything. You decide what goes, box by box. There is no automatic detection of names, card numbers or addresses, and a tool offering that would still miss things while sounding confident about whatever it did catch.

It also only deals with what you cover. If the name you redacted on page four is also in a running header on page nine, in a footnote, or spelled slightly differently in an index, those copies are still in the document. Search the file for the information first, then mark every place it turns up.

Two things outside the pages deserve a thought as well. A file name travels with the document, so redacting a page and sending it as smith-dispute-final.pdf gives the name away regardless. And wherever the original came from, an email or a shared folder, still has the original in it.

Nothing leaves your device

The document is opened, rendered, rebuilt and verified in your browser. No upload, no copy on a server, no temporary file on anybody else machine, which for a document you are in the middle of redacting is the entire reason to use something built this way.

That also sets the ceiling. Drawing pages at 300 DPI takes memory, and a very long document at the highest setting can exhaust a modest phone. The limit here is 50 MB, lower than the other PDF tools on this site, because this one has to redraw pages rather than rearrange them.

Limits

50 MB per file.
Pages you mark become images, so they lose selectable and searchable text.
Those pages come out larger, sometimes several times larger.
Bookmarks, attachments and form fields do not survive the rebuild.
Text you did not cover is untouched, including the same words on other pages.
Password-protected PDFs have to be unlocked first.
On a scanned page there is no text layer to search, so the check can only confirm the page was rebuilt.
This is not a compliance product. Whether a redaction satisfies a court, a regulator or an employer is not something this page can tell you.

Frequently asked questions

Can someone get the text back from under the boxes?: Not from the pages you marked. Those pages are rebuilt from a flattened image with the boxes already painted in, and the original page is never copied into the new file, so there is no hidden layer to recover. The finished file is also searched for the text that was under your boxes, and if any of it is still findable you get an error rather than a file.
Why can I no longer select the text on that page?: Because the page is now a picture. Taking text off a page without rebuilding it is exactly what leaves recoverable copies behind, so losing the text layer on a redacted page is the price of the redaction being real. Pages you did not mark keep theirs.
Why is my file so much bigger?: An image of a page takes far more space than the text on it, and only the pages you marked are affected. If size matters more than crispness, drop the quality setting to 150 DPI, which is enough for reading on a screen.
Does it remove my name from the document properties?: Yes. Title, author, subject, keywords and producer are all cleared in the output, because those fields often carry a name or an internal path that nobody intended to send on.
It refused to give me a file. What went wrong?: The check after the rebuild still found text that your boxes were meant to remove, so nothing was handed over. This is deliberate: a file that looks redacted and is not would be worse than an error message. Make sure the boxes fully cover the text, including any part of a line that pokes out of the edge, and try again.
Can I redact a signature or a photograph?: Yes. The boxes are painted into the pixels of the page, so anything under one goes, whether it was text, a signature, a stamp or part of an image.