How to redact a PDF without uploading it

Of all the PDF jobs, redaction is the strangest one to do by upload. The reason you are redacting is that the document contains something sensitive, and the first step of the popular tools is to send the whole unredacted thing to a server. Then there is the second problem, which is that a lot of what gets called redaction isn't.

A black rectangle is not a redaction

In 2019, a court filing by Paul Manafort's lawyers was “redacted” with black boxes. Reporters selected the blacked-out passages, pasted them into a text editor, and read them. In 2020 the same thing happened with the unsealed Ghislaine Maxwell deposition. Both were drawn over, not removed: the text sat in the file underneath the rectangle, exactly where a text extractor expects to find it.

Real redaction destroys the content. The reliable way to do that for a page is to rebuild it as an image with the marked areas painted out, so the words no longer exist in any layer, and then to check the result rather than assume it.

Redacting in the browser, without an upload

1. 1Open Redact PDF and choose the file. The first page renders in the tab.
2. 2Type the name, number or phrase to remove and click Mark all matches. Every occurrence on every page gets a box. Repeat for each term; the terms are remembered for the verification step.
3. 3For anything search can't see, drag on the page to draw a box. Click a box to remove it.
4. 4Pick a mode. Rebuild only marked pages keeps selectable text on every untouched page; Flatten rebuilds the whole document, which is what you want for scans or when you'd rather remove every text layer.
5. 5Click Redact & verify. The tool builds the output, re-opens it as if it were a stranger's file, and shows the report: zero extractable text on redacted pages, no surviving search terms, metadata stripped. If a term survived on a page you didn't mark, the report names the page.
6. 6Download. Then open the file and try to select the redacted area yourself. Nothing selects, because nothing is there.

Open Redact PDF →

What the verifier does and doesn't prove

The report proves the text layer is gone and your terms are absent from the file, and it confirms the metadata was stripped, which matters because author names and revision history leak through metadata more often than through page content. What it cannot judge is whether you marked the right things: if a phone number is written in a way you didn't search for, it is still there, in plain sight, on a page the report calls clean. Search for every variant you can think of, and read the output once before it leaves your hands.

Scans and photographed documents

A scanned PDF is pictures of pages. There is no text to search, so Mark all matches finds nothing and says so. Draw the boxes by hand, use Flatten, and check the output visually. The upside is that a scan has no hidden text layer to leak in the first place, unless someone ran OCR on it, in which case Flatten removes that too.

Why local matters more here than anywhere

An upload-based redaction service receives the unredacted document. Whatever happens next, retention windows, backups, logs, a breach, the sensitive version has already left your control. On PDFAirlock the file is opened, rebuilt and verified by code running inside your tab. You can watch the Network tab stay empty, you can turn off your wifi first, and you can read the page's Content-Security-Policy, which forbids the browser from contacting any host but the site itself. The privacy page lays out all three checks, and the note for law firms covers the professional-duty angle.