How to Redact a PDF Properly, Not Just Cover It Up

Why doesn't drawing a black box redact a PDF?

A PDF page is built in layers. The text you see is stored as actual characters with positions and a font, and images, shapes and annotations are drawn on top of or beneath it. When you add a black rectangle with a markup tool, a comment tool or a generic editor, you are adding one more layer. The text underneath is untouched. It is simply painted over.

Anyone who receives the file can often recover that text in seconds: by dragging the cursor across the black area and copying, by using the viewer's search function, by deleting or moving the rectangle in an editor, or by extracting the text with any conversion tool. This failure is well known among lawyers, journalists and records officers precisely because it is so easy to make, and it has repeatedly caused sensitive information to leak from documents that looked perfectly redacted.

The same problem applies to highlighting text in black, setting a black background colour, using white-out, or covering text with an inserted image. If the original characters remain in the file, the redaction is cosmetic.

• Black rectangle annotations: text remains underneath.
• Black highlight or text background: text remains and is often still selectable.
• White-out or pasted shapes: text remains underneath.
• Cropping a page: the cropped-away area is usually still in the file.

What does proper PDF redaction actually do?

Real redaction removes the underlying content, not just its visibility. After redacting, the characters, image pixels and any other data in the marked area must no longer exist in the file. The black box that remains is just a marker telling the reader that something was removed.

There are two broad ways tools achieve this. Some professional editors rewrite the page content, surgically deleting the text objects and image regions inside each redaction area while keeping the rest of the page as live text. Others flatten the page: they render it to an image with the black boxes burned in, and replace the original page with that image. Either way, the original data is gone.

Good redaction also thinks beyond the visible page. Metadata such as author names and document titles, hidden layers, comments, form field values, bookmarks and attached files can all contain information you meant to remove. Before sharing a sensitive file, consider whether any of those could give away what you just blacked out.

How do you redact a PDF with pdfdots?

Our Redact PDF tool runs entirely in your browser. For redaction that matters more than usual: the document you are redacting is by definition sensitive, and with pdfdots it is never uploaded to a server. There is no account and no watermark.

When you apply the redactions, every page that contains at least one redaction box is flattened: we render that page to an image with the black boxes drawn in, and replace the original page with the image. The text, vector graphics and any content under the boxes on that page no longer exist as data in the downloaded file. Pages you did not redact are left as they were.

We picked flattening because it is the approach we can guarantee. Surgically deleting text from a PDF's content stream is complex, and a subtle bug can leave fragments behind. Rendering the page to pixels leaves nothing to recover under the box.

• Open Redact PDF and drop in your document.
• Drag boxes over every name, number, signature or passage that must go.
• Page through the whole document, including appendices and footers.
• Apply the redaction and download the new file.
• Verify the result using the checks in the next section.

How can you verify that a redaction worked?

Never assume. Whatever tool you use, including ours, test the downloaded file before it leaves your hands. These checks take a couple of minutes and catch almost every cosmetic redaction.

Do these checks on the final file you are going to send, not on the working copy. It is surprisingly common to verify one version and then attach another.

• Select and copy: drag your cursor across each black box, copy and paste into a plain text editor. Nothing from the hidden content should appear.
• Search: use the viewer's find function to search for a redacted word, name or number. It should return no matches.
• Try a second viewer: open the file in a different PDF reader or browser to be sure nothing renders differently.
• Convert it: run the file through a PDF to text or PDF to Word conversion and check the output for the removed content.
• Check the properties: look at the document properties for title, author and subject fields that may reveal names.

What are the trade-offs of flattening redacted pages?

Because redacted pages become images, the text on those pages is no longer selectable or searchable, including text you did not redact. A reader cannot copy a paragraph from a flattened page, and screen readers cannot read it aloud. File size for those pages can also increase compared with the original vector text.

For most redaction jobs we think this is the right trade: the recipient of a redacted document is reading it, not editing it, and certainty that the data is gone matters more than copyable text. If you need the redacted document to be searchable again, you can run OCR on the result. OCR only recognises what is visible on the page image, so it cannot bring back the redacted content.

One more practical tip: redact generously. Leave a little margin around each item, and remember that the length of a black box can itself hint at what was removed, such as a short name versus a long one. When that matters, make boxes a consistent size.

What should you redact besides the obvious?

People usually remember the headline items and miss the supporting details that identify the same person or account. Before applying, scan the document for these often-forgotten places.

• Headers, footers and page margins that repeat names or reference numbers.
• Signatures, initials and handwritten notes on scanned pages.
• Email addresses, phone numbers and account numbers inside tables.
• Images and screenshots containing names, faces or visible data.
• Filenames, document titles and metadata that describe the content.

Frequently asked questions