Why redactions fail
The failure mode is simple and common: instead of deleting the sensitive text, the author draws an opaque rectangle on top of it. The page looks redacted. The text layer is still there, and a reader can select it, copy it, search it or extract it with a script.
Researchers at Essex Software audited 72 publicly indexed PDFs whose titles or URLs contained "redacted"; 63 could be audited. 11 of the 63 (17%) had selectable text physically present under the redaction rectangles. The rate was highest for US federal agency files: nearly one in three.
| Finding (n = 63 audited PDFs) | Documents | Share | Severity |
|---|---|---|---|
| Selectable text under redaction rectangles | 11 | 17.5% | Critical |
| Any identifying metadata leaked | 59 | 94% | Warning |
| Producer field set (software trail) | 49 | 77.8% | Warning |
| XMP metadata stream present | 56 | 88.9% | Warning |
| Document title set | 42 | 66.7% | Warning |
| Author field set | 38 | 60.3% | Warning |
| Unapplied /Redact annotations | 0 | 0% | Critical |
Source: Essex Software, "We audited 72 'redacted' PDFs" (June 2026). The study defines critical leaks as selectable text under a visible opaque rectangle.
The 60-second self-check
You do not need a special tool to catch the most common failure. Do this before you file, publish or send the PDF.
Select across the black boxes
Open the PDF, drag the text cursor across a redacted area, and copy. Paste into a plain-text editor. If the hidden words appear, the text layer is still present. This is the single highest-value check.
Search for the secret words
Use the PDF reader's search for the exact names, account numbers or phrases you meant to hide. A hit inside a redacted region means the text is still extractable.
Extract all text
Tools such as pdftotext file.pdf - (Poppler) dump the text layer. Search the output for the terms that should not be there. This catches text that is invisible or covered.
Inspect the metadata
Author, title, producer, subject, keywords, XMP, bookmarks and attachments all travel with the file. Use the PDF metadata viewer or the all-format viewer to see what is there.
One caveat: a clean copy-paste test does not prove the file is safe. Text can also hide in invisible layers, annotations, attachments, form fields, OCR layers or incremental saves. The checks above catch the dominant failure mode, not every possible one.
What a real redaction requires
True redaction is a destructive operation: the content must be removed, not covered. A proper tool will delete the text and images in the marked area, then rewrite the page so the removed content no longer exists in the file.
After removing content, the document still needs a cleanup pass: metadata, XMP, bookmarks, annotations, attachments and hidden layers can all carry the same secret. The Essex audit found zero documents with unapplied /Redact annotations; the failures were almost entirely "shapes over live text".
- Delete, do not cover. Use a real redaction tool, not the rectangle or highlighter tool.
- Flatten or rebuild the page if the tool offers it, then verify by extracting text again.
- Strip the metadata separately: the PDF metadata remover handles Info fields, XMP and attachments, but it is not a redaction tool.
- Check bookmarks and attachments. A bookmark title can quote the redacted section; an attachment can be the original file.
- Re-run the self-check on the output. Never trust the tool's "redacted" label without testing the final bytes.
What a PDF redaction checker should tell you
A useful checker does not just look at pixels. It reads the document structure and reports what is still recoverable. This is what we are building, locally in the browser:
| Check | What it means | Why it matters |
|---|---|---|
| Selectable text under shapes | Text is still present under an opaque rectangle | The critical failure mode; can be copied or extracted |
| Hidden text outside the crop | Text positioned off-page or clipped | Often survives viewers and extraction |
| Metadata and XMP | Author, title, producer, custom fields | Identity leaks even when the text is gone |
| Bookmarks / annotations / attachments | Titles, comments and embedded files | Can quote or contain the redacted content |
| Layers / optional content | Hidden layers that can be switched on | Rare, but a known carrier |
| Report | A shareable summary of findings | Evidence that the check was done |
If the problem is only metadata
Redaction failure and metadata leakage are related but different. If the visible text is genuinely gone and you only need to remove author, title, XMP and attachments, the PDF metadata remover does that locally in the browser. If you need to inspect first, the PDF metadata viewer and the all-format viewer show what is inside without uploading the file.
Limits
- No checker in Docfy yet. This page is a guide; the early-access list is for the local checker when it ships.
- Self-checks are not certification. Passing the copy-paste test does not guarantee the file is clean.
- Not legal advice. Whether a redaction is sufficient for a court, regulator or contract depends on the rules that apply to you.
- Scanned documents differ. If the PDF is an image-only scan, the failure mode is different: text may be baked into pixels or added by OCR, and only visual inspection catches it.
Questions people ask
What is a PDF redaction checker?
A tool that checks whether the text under a PDF's black boxes is actually gone, and whether the file still carries identifying metadata, bookmarks, attachments or hidden layers. It answers "did my redaction work?", not "is this document legally compliant?".
Why do redactions fail if the text looks blacked out?
Because drawing a black rectangle does not remove the text underneath it. The text remains in the content stream and can often be selected, copied, searched or extracted. A real redaction deletes the content and then replaces the page or image.
How do I check a PDF for hidden text?
Open the PDF and drag-select across the black boxes; copy and paste into a plain-text editor. If the hidden words appear, the text layer is still there. Also search for the redacted terms and inspect metadata, bookmarks and attachments.
Does removing metadata fix a failed redaction?
No. Metadata removal clears fields like Author, Producer, XMP and bookmarks, but it does not remove text that is still present in the page content. Redaction and metadata cleaning are different jobs.
Is flattening or printing to PDF enough?
Sometimes. Rasterising the page to an image and rebuilding the PDF removes the text layer, but it can also destroy accessibility, searchability and quality. If you flatten, verify the output by extracting text again.
Is there a local PDF redaction checker in Docfy?
Not yet. The plan is a browser-local checker that lists selectable text under shapes, metadata fields, bookmarks, annotations and attachments, and produces a report you can keep. Join the early-access list if you want it when it ships.
Is this legal advice?
No. This guide explains a document-integrity check. Whether a redaction satisfies a court, regulator or counterparty depends on the rules that apply to you; consult a lawyer for that question.