Why black boxes on a PDF often hide nothing

A document can look perfectly redacted on screen and still hand over every name it was meant to hide. The reason is that a PDF stores what you see and what you can copy in separate layers.

1. The text is still under the box

Most PDF editors let you draw a filled rectangle. That rectangle is a new drawing placed on top of the page; the characters underneath stay in the file. Anyone who selects the area and copies it, or opens the file in a text extractor, reads the original. Real redaction deletes the characters from the content stream and only then paints the box, so there is nothing left to copy.

A quick check: open the redacted file, press select all, paste into a plain text editor and search for one of the names you removed. If it appears, the redaction failed.

2. Metadata travels with the file

Author names, the original file name, revision history and the software used are stored in the document properties and in XMP metadata. A file called Rossi_divorce_final.pdf with the author field set to a colleague's name tells the reader a lot before they open a single page. Clear the document properties and the XMP packet before sending.

3. Scans hide text inside an image

A scanned page is a picture, so the text search in step 1 finds nothing, and a tool that only redacts text will miss it too. If the scan was run through OCR, there is also an invisible text layer on top of the image that can still be copied. A scanned document needs both: the pixels under each box burned out of the image and the OCR text removed.

4. The same name in places you did not look

Names reappear in headers and footers, in bookmarks, in form fields, in annotations and comments, and in attachments embedded in the PDF. Identifiers also hide in plain sight: an IBAN or a national ID number split across two lines is easy to miss by eye.

A short checklist before sending

Copy all text from the exported file and search it for every value you removed.
Check document properties and remove the XMP metadata.
For scans, confirm the boxes are part of the image, not a layer above it.
Flatten or remove form fields, comments and attachments.
Search the whole file for each identifier, not just the page where you saw it.

Tools built for redaction, such as Redactic, remove the text, burn the boxes into the page and strip metadata in one export, which takes the manual checks off your list. Whatever tool you use, the copy and search test above takes ten seconds and catches the most common failure.

More on detecting personal data before it leaves the office: data redaction guide for GDPR and AI.