Sometimes. If a black box was only drawn over the text, the text is still in the file and can be copied or searched. If the redaction removed the underlying content, that file cannot simply be reversed, but duplicates, hidden data or clues in the remaining document can still reveal information. The two files can look identical on screen: inspect the saved file, not just its appearance.
A black rectangle is evidence of intent, not of removal.
Try two files that look the same
Download the black-box sample and the removed-text comparison. Both are one-page fictional expense records showing receipt H-028 and USD 480.00. Neither contains customer data.
- In
covered-text.pdf, the fictional name Casey Roe is still present beneath the black rectangle. Try copying the page or searching for the name. - In
removed-text.pdf, the name was omitted when the PDF was generated. The same black rectangle is drawn, but there is no name underneath it to extract. This comparison demonstrates absence in this generated file, not a tested redaction editor or a DocHush desktop export.
Open the free browser checker’s example instructions, enter Casey Roe as a term that must be gone, and check each downloaded file. The tool page records the tested version, settings and observed results. Match its reported SHA-256 to the sample if you need to identify the exact file.
If your result differs, follow the sample troubleshooting steps: check the original download’s hash, match the settings, then read the individual findings and coverage. “Risk found” need not be a match for the name you searched for, and “Could not verify” is not a pass.
The files deliberately isolate a simple text-cover case. They do not test scans, OCR, attachments, incremental saves or inference from context. Keep those limits in view when checking a different PDF.
When a redaction can be reversed
The first distinction is whether content was removed or merely hidden. The ICO’s technical example shows text surviving a black highlight and PDF export, then being recovered by copying it into a text editor. ICO — How to disclose information safely, paragraphs 58–65
- A shape, highlight or comment on top of the text. A cover can leave text in the page, available to selection, copying, search or extraction. Removing an annotation may also reveal it.
- Text made invisible. Changing text to the background colour or cropping it out of view can leave it in the file.
- Hidden information elsewhere in the file. Document properties, comments, attachments, bookmarks and form field values can carry the same details. A PDF saved with incremental updates can also contain its earlier versions, including the unredacted page.
- A text layer accompanying an image. OCR can add searchable text to a scanned page. Changing the image alone does not necessarily remove that text.
- Glyph positions. A 2022 study by Bland, Iyer and Levchenko recovered names using character-position information left after the text itself was removed. Its findings describe the versions studied, not a test of current tools. Story Beyond the Eye
- Context. The width of a redaction and the surrounding words can help someone infer the missing value. Review whether a wider passage needs removal while preserving what the recipient is meant to receive.
Five checks for the saved PDF
Run these on the exact copy you intend to share. Keep a list of values that must be removed and information that should remain. Use a second viewer as an independent check; if a tool cannot open or inspect the file, record that limitation instead of treating the missing result as a pass.
- Select and copy around each box. Drag across the redacted area and a little beyond it, paste into a plain-text editor, and read what arrives. This can expose text left under a cover. Failure to select it does not establish removal.
- Search for known values. Search for the names, full numbers and distinctive fragments that must be removed. Try relevant spacing and spelling variants. Inspect each hit: a shared fragment may belong to a different field approved to remain.
- Extract text with a second program. Use a viewer’s text export or another text extractor, then search that output. This may reveal hidden text missed by the first viewer. Empty, garbled or incomplete output is a coverage limitation, not evidence of successful redaction.
- Inspect hidden data. Review properties, comments, attachment listings, bookmarks, form values and hidden layers with a tool that exposes them. Use the editor’s hidden-information removal step where available, then inspect the saved copy again. Visible panels alone do not audit earlier revisions or every stored object; the browser checker reports its coverage for these separately.
- Read every page visually. Zoom in on marked areas for partial letters or uncovered digits, then check unmarked content, headers, footers and continuation pages for missed occurrences. A bank statement, for example, repeats the account number in each page header; the bank statement guide shows that check. Remove labels only when your sharing decision requires it. Compare the page count and order with the intended output.
For a file you produced, add one more comparison: put the original and the redacted copy side by side and confirm that everything the recipient needs is still readable.
Decide what to do with a finding
A finding tells you what the check encountered. Your sharing requirements determine whether that information should remain. Read the finding and the coverage together before deciding what to do next.
| What you find | What to do next |
|---|---|
| A value that must be removed is still present. | Correct the page or hidden object in your working file, export a new copy, and check it for other occurrences. |
| A flagged value is approved to remain. | Confirm the field and its location against the sharing brief. A correct pattern match is not a decision to delete. |
| A page or object could not be verified. | Read the error and coverage. Use a supported, readable copy or another suitable inspection method. Visually review images; hold the file if a required check remains unresolved. |
| An attachment’s contents were not checked. | Remove it if it is outside the delivery scope. Otherwise inspect and redact it separately, then verify the final package again. |
For the fictional expense example above, suppose the recipient needs receipt H-028 and USD 480.00, but not Casey Roe. A match for the name would require correction; a flag on a field approved to remain would require review of that decision, not automatic deletion. After correcting the file, check that the receipt and amount are still readable and that the name is absent within the checks you ran.
What a clean result does and does not prove
No matches means those tools did not find the supplied terms in the content they could read. It does not establish that every copy is absent: extraction may be incomplete, wording may differ, and image text is outside a text-only search. These checks also do not test glyph-position leakage or rule out inference from context.
Keep a review record tied to the final filename or SHA-256: pages checked, tools and versions, search terms, date, finding locations, decisions to retain information, corrections and unresolved coverage. Store sensitive terms with the protected review materials. Recheck if you change or resave the PDF after review.
Correct a failed redaction and check the tradeoff
Return to the working file, apply content-removing redactions and handle hidden information. One output option is a new PDF built only from page images whose sensitive pixels have been replaced. This discards the original text objects and their precise positioning data. Simply printing to PDF, flattening annotations or placing an image above the original content does not establish that this happened.
Image-only output still preserves visible spacing and surrounding context. The glyph-position study explains that approximate redaction width can leak even after rasterization. It also loses searchable text and the original accessibility structure, links, interactive fields and digital signatures. Confirm the recipient can use that format before choosing it. Story Beyond the Eye, sections 3 and 7.1
Keep the original separately, save the redacted copy under a distinct name, and verify the saved copy with the checks above before it leaves your computer. Our guides on redacting without Adobe Acrobat and redacting scanned PDFs cover tool choice and image-only pages in more detail.
Where DocHush fits
The free DocHush PDF Redaction Checker can inspect the saved PDF now. It looks for text under covers, invisible text, metadata, attachment listings, bookmarks, form values, hidden layers, earlier versions and your search terms. PDF data, search terms and results stay in your browser. Read its coverage statement: it uses one parser, does not run OCR or open attachment contents, and does not exhaustively audit unreferenced objects.
In the DocHush workflow, Verify reopens the exported file and checks its structure, page consistency and the regions you approved within a stated coverage scope. Read what completed and what could not be verified. These checks do not decide what should be removed or establish that unmarked areas contain no sensitive information; they provide a record of the checks performed, not a certificate.
The export workflow rebuilds each output page from an image with the approved regions burned in, without carrying over the original text layer. Read the resulting images for missed content and contextual clues as well as the Verify report. Download DocHush for Windows.
Frequently asked questions
Can you remove redaction from a PDF?
A removable cover may reveal text that was never deleted. Removing a box cannot restore content that is absent from the file, but other copies or clues elsewhere can still disclose it. That is why both the marked regions and the rest of the file need review.
Is an empty text search enough for an image-only PDF?
No. The words may still be readable in the page image. Review the pixels on every page and confirm the export replaced the intended regions. A text-only checker cannot make that judgment.
Is a black box the same as redaction?
The same-looking box can accompany either a cover-up or removed content. Its appearance alone does not tell you which; use the saved-file checks and their stated limits.