BASALT · JOURNAL

Removing a black box from a PDF, and what it means that you can

2026-08-07 · how to remove black box from pdf

People search how to remove black box from PDF pages for two different reasons, and the honest answer is the same for both: if the box is an annotation, it deletes in a single operation and the text underneath is completely intact. That is not a trick or an exploit, it is the documented behavior of a comment object, and it is precisely why drawing a box is not redaction.

This article is written for testing your own documents. If you drew a box over a client name and want to know whether the file is safe to send, the procedure below is the test. Run it on a copy of your own file and act on the result.

Two reasons people search how to remove black box from PDF

The first is that you received a document with a box on it and cannot read a value you need. The second is that you drew the box yourself and want to know whether it protected anything.

Those two situations look identical from the file's perspective, which is the whole problem. A file where the words were excised and a file where a rectangle was painted over them render the same black bar at every zoom level. Nothing about the appearance distinguishes them.

The removal mechanism is described here because you need it to audit your own work, not to open other people's confidential material. Defeating a third party's redaction of information you were not given is a different activity with different consequences, and this is not a guide for it.

Why the box comes off so easily

Most black boxes in PDFs are annotations. Markup tools in Preview, in Mail, and in most annotation-first apps create an object in the page's /Annots array: a dictionary with a rectangle, a subtype such as /Square, a color, and an appearance stream.

Annotations exist to be added and removed. They sit alongside the page content rather than inside it, and stripping them is a structural edit that any conforming library performs without touching a single text operator. When the annotation goes, the page underneath renders exactly as it did before, because it was never modified.

The second common case is a box drawn into the content stream itself, as fill operators after the text. Deleting the annotation does nothing there, but the words are still present in the Tj and TJ operators regardless of what is painted on top, so extraction returns them anyway.

Testing your own file

Work on a copy. Install Poppler and qpdf with Homebrew, then start with the cheapest check.

pdftotext -layout mydoc.pdf - | grep -i "confidential-term"

If the term you thought you removed comes back, the box was cosmetic and the file is not safe. A clean result proves less than it looks like, because fonts without a usable /ToUnicode map can hold the original character codes while producing empty output.

Expand the file so the structure is readable as plain text and look for annotations near the box.

qpdf --qdf --object-streams=disable mydoc.pdf expanded.pdf
strings expanded.pdf | grep -i "confidential-term"

In expanded.pdf, search for /Annots and for /Subtype /Square. A rectangle in an annotation dictionary, with the text still present in the content stream below it, is the exact signature of a document that will unredact itself for anyone who deletes comments.

If the redaction covered a scan, extract the images and open them directly, since a page-level box does nothing to the pixels inside an embedded image object.

pdfimages -all mydoc.pdf ./extracted/img

Finally, count generations, since more than one end-of-file marker means earlier revisions are still in the file and one of them may be the page before you covered anything.

strings mydoc.pdf | grep -c "%%EOF"

A longer version of this sequence, with the metadata and attachment checks included, is in how to verify a PDF redaction. If the terms come back at any stage, the situation is the one described in your redacted PDF still shows the text.

What to do when the box fails the test

Go back to the original document rather than repairing the marked-up copy. A repaired copy can still carry the pre-redaction page in its update chain, and it will have accumulated metadata and thumbnails from the failed attempt.

Real removal means editing the content stream. The text-showing operators are parsed into individual glyph runs with real page coordinates, accounting for the text matrix, font size, horizontal scaling, character and word spacing, and the kerning numbers inside a TJ array. Covered glyphs are excised and the survivors re-emitted at their original positions.

Where the content is pixels, the pixel data under the mark is overwritten and the image re-encoded, so the original DCTDecode stream cannot be pulled out whole. Where the page carries an invisible OCR layer drawn in text rendering mode 3, that layer has to be excised at the same coordinates as the visible mark.

Then the file is written as a single generation with a fresh cross reference table, and the carriers that are not on any page are cleared in the same pass: the document information dictionary, the XMP packet, embedded attachments, optional content groups, annotation /Contents strings, and stale thumbnails. What a PDF still carries after you think you cleaned it lists the full set.