If you've drawn a black rectangle over a name, an account number, or a paragraph in a PDF and you're about to send that file to someone, stop for a second. The rectangle is a drawing. It sits on top of the page like a sticky note. The text underneath is still there, in the exact same position, stored the same way it always was — and getting it back usually takes one keystroke: select all, then copy.
Why a Black Box Doesn't Remove Anything
A PDF page isn't a flat picture. It's a list of independent objects — text, lines, images, shapes — each with its own position in a stacking order. "Blacking out" a sentence with a rectangle tool just adds one more object: a shape placed above the text. The text itself is untouched, still real character codes sitting in the page's content stream. It still shows up when you:
- Select the area with the mouse and copy it — many viewers happily select text that's visually covered by something drawn on top.
- Run Ctrl+F / Cmd+F and search for the word you thought you hid.
- Open the file in any tool that extracts text programmatically — there are dozens, and most are one line of code.
- Delete or drag the black rectangle out of the way in an editor, exposing the original text underneath, fully intact.
This isn't a theoretical bug. Text hidden behind a drawn box, later recovered by copying, searching, or nudging the box aside, is a well-documented way information has leaked from court filings, government releases, and corporate reports. A black rectangle equals coverage, not removal.
Why We Call It "Black Box," Not "Redact"
PDFCraft's Sign & Edit tool includes a black rectangle tool, deliberately labeled "Black Box," not "Redact," with a warning shown whenever you use it. "Redact" implies the content is gone; it isn't. Sign & Edit draws overlays — typed text, a signature, freehand lines, a highlighter, a filled box — on top of the page; it doesn't touch or delete text that's already there. That's true of most browser-based annotation tools; the difference is whether the tool admits it. If a tool calls a rectangle "redaction" without saying the text survives, be skeptical of everything else it says.
Four Ways to Actually Remove Content, Ranked by Reliability
There's no single button that handles every case; which method fits depends on how sensitive the content is and what's at stake if it leaks.
| Method | Underlying text removed? | Stays searchable/selectable? | Effort | Best for |
|---|---|---|---|---|
| Delete the whole page | Yes — page and content stream gone | No — page no longer exists | Very low | Content confined to one or more full pages |
| Flatten page to an image | Yes — page becomes a picture | No — whole page loses selectable text | Low-medium | One sensitive line on a page you still need |
| Fix it at the source, re-export | Yes, cleanest result | Yes, if you want it to be | Medium | You still have the original Word/Docs file |
| Dedicated redaction software | Yes, if the tool strips content objects | No for redacted areas, yes elsewhere | High | Legal filings, FOIA requests, regulatory work |
1. Delete the page
PDFCraft's Delete Pages tool builds a new document from only the pages you keep, so a fully sensitive page carries nothing into the output. See organizing PDF pages for how it fits alongside reordering.
2. Flatten the page to an image
The reliable option when only part of a page is sensitive: the whole page becomes one picture with no underlying text objects, covered step by step below.
3. Fix it at the source
If you still have the source file, delete the content there and export fresh — nothing old to carry over, and the only method here that preserves searchable text elsewhere on the page.
4. Dedicated redaction software for high-stakes work
For legal discovery or records requests, use software built specifically for redaction that strips content objects rather than drawing over them, and verify the output yourself (see below). PDFCraft doesn't offer that; a purpose-built desktop tool is the right choice here.
Step by Step: Flatten a Page So the Text Is Gone for Good
This is the practical middle ground — no original source file required, and unlike a black box it genuinely destroys the text layer.
- Open PDF to Images and convert the document to PNG or JPG. Use 300 DPI for small text that needs to stay legible; 150 DPI suits a quick review copy. Our guide on converting between PDF and images covers the trade-offs.
- Open the sensitive image(s) in an editor and cover the area by cropping it out or filling it solid — never blur or pixelate, since both are only partially reversible for short strings like account numbers. A flat picture has no text hiding underneath, because images contain no text objects at all.
- Open Images to PDF, upload the page images in order, and rebuild the PDF.
- Verify (next section) before you send it anywhere.
The trade-off: the whole page you flatten loses its text layer, not just the redacted part — unselectable, unsearchable, unreadable by screen readers. Flatten only the page that needs it.
How to Verify Removal Actually Worked
Do this every time, regardless of method. It takes two minutes and it's the step people skip.
- Try to select the text. Click and drag over where the sensitive content was. If a cursor appears and something highlights, it's still there.
- Search for it. Ctrl+F / Cmd+F for the exact word, name, or number you removed. A flattened page should return zero results.
- Copy the page into a plain text editor. Paste into Notepad, TextEdit, or any plain-text app. If the sensitive line pastes in as text, it's still in the file no matter how it looks.
- Open it in a second, unrelated PDF viewer if you can — a different renderer sometimes displays a document differently, catching viewer-specific quirks.
If any of these bring the content back, go back and delete the page or flatten it — don't add another box on top.
Metadata and Hidden Content People Forget
Even a properly flattened, verified page can leak information that never appears on the page itself: the file name (rename it before sending), document properties like Title and Author (visible under File > Properties in most viewers, often carrying the real author's name), tracked-change remnants that can survive a "print to PDF" step, and Optional Content Groups — hidden layers, common in CAD or design exports, that a viewer's Layers panel can switch back on.
The PDF to Images / Images to PDF round trip above sidesteps annotations and layers on that page, since none of it survives being rendered to a flat picture. It also drops the original Title and Author fields, because Images to PDF builds a brand-new document rather than editing yours — pdf-lib writes its own generic Producer/Creator instead. It won't fix your file name, though; rename it yourself.
Heads up: None of this makes a sharing decision safe by itself. Client-side tools like PDFCraft never upload your file to a server — see our guide on choosing safe online PDF tools for confidential documents for why that matters — but a correctly redacted PDF sent to the wrong email address is still a leak. And a flattened page is no longer accessible to screen readers, so flatten only the page that needs it, not the whole document.
Frequently Asked Questions
I drew a black box and then copied the page into a Word document — did that fix it?
No. Pasting a PDF page into Word usually keeps the same layered structure — a picture or text run with a shape on top — so the text can still be extracted the same way. Only converting to a flat image (screenshot, or PDF to Images) collapses the layers into one.
If I flatten a page to an image, can I make it searchable again afterward?
Only by running OCR (optical character recognition), which builds a brand-new text layer from what the engine reads off the picture — it won't include whatever you removed before flattening. PDFCraft doesn't include an OCR tool, so a flattened page stays non-searchable otherwise.
I used a black highlight in Word or Google Docs before exporting to PDF — is that safe?
It depends on the export. Some paths genuinely delete the underlying text for certain "redaction" or highlight styles; others just format it with a black background and export it unchanged, reproducing the same problem. Always run the verification steps above on the exported PDF before trusting it.
Can a PDF hide information in a way that isn't a black box at all?
Yes. Optional Content Groups (layers) can exist in a file but be switched off by default, showing only if someone opens the Layers panel in a full-featured viewer. This is more common in PDFs from design or CAD software than typical office documents.
Does deleting a page with PDFCraft leave any trace of it behind?
No. Delete Pages builds a new document from only the pages you keep, so the removed page's content stream is never written into the output. It also refuses to delete every page, since a PDF must keep at least one. It can't account for history from before the file reached you, though: earlier revision data from other editors could still exist independent of PDFCraft.
Is there any single tool that reliably redacts a PDF for legal use?
For genuinely high-stakes redaction — court filings, FOIA responses, regulated data — use desktop software built to strip the underlying content object rather than draw over it, then verify the result using the checks above regardless of what the software claims.