Redact PDF

The mistake that keeps leaking documents

A PDF is not a picture. It is a set of instructions describing where each glyph goes, and drawing a black rectangle adds one more instruction on top of them. The text underneath is untouched: select over the black box, copy, and paste into a text editor, and there it is.

This is not a theoretical failure. It has exposed the names of confidential informants in court filings, revealed unredacted passages in government reports, and disclosed negotiated figures in contracts released to the press. In several well-publicised cases the redaction was made in a professional PDF editor by people who believed the black boxes were doing the work.

The same trap catches a common workaround: highlighting text in black in a word processor before exporting to PDF. The characters are still characters; they simply have a black background and black foreground.

How this tool actually removes the text

Each page is rendered to an image at the resolution you choose — the same rendering a PDF reader performs to display it. The areas you marked are painted out on that image, and a new document is built from the results.

Because the new page is an image, the original text objects are not carried over. They are not hidden, not covered and not merely obscured: they are absent from the file. Selecting over the redacted area produces nothing, because there is nothing there to select.

This also removes any metadata attached to the original page content, along with annotations and form data, which is usually what you want when the point of the exercise is to publish a document safely.

The trade-off, stated plainly

Rasterising the pages costs you real things, and it would be dishonest to bury them. The whole document stops being searchable. Text can no longer be selected or copied anywhere in it, not just in the redacted regions. Screen readers can no longer read it, which is an accessibility regression. Vector graphics are flattened, so zooming in eventually shows pixels.

File size usually increases for a text-heavy document, because a page of prose stored as an image is far larger than the same page stored as glyph instructions.

For most redaction work this is an acceptable price, and it is the same trade-off made by the traditional method of printing a document, marking it with a black pen, and scanning it back in — a method that is entirely secure precisely because it destroys the underlying data.

If you need redaction that preserves searchable text elsewhere in the document, you need software that can edit the content stream and remove specific text objects. That is a desktop-class capability, and any browser tool claiming to do it is worth treating sceptically.

Checking your work before you send it

Always verify. Open the downloaded file, try to select text over a redacted area, and confirm nothing is selectable. Use your reader's search function to look for a term you removed — if it finds nothing, the removal worked.

Two things this tool cannot do for you. It cannot find sensitive information you did not mark, so read every page rather than only the ones you remember. And it cannot redact what is not visible — information hidden in document metadata, in comments, or in earlier revisions saved inside the file.

Be alert to inference as well. Redacting a name but leaving a job title, a date and a location can identify someone just as effectively. Redaction is an editorial exercise as much as a technical one.