Loading page
About this tool
Inspect redacted PDFs and images locally for exposed text, risky annotations, metadata, and visual warning signs before sharing.
The Redaction Proof Checker opens a redacted PDF or image in your browser and looks for the evidence that a redaction failed — text still extractable under a black box, redaction annotations that were drawn but never applied, embedded attachments, document scripts and identifying metadata. PDFs are parsed with PDF.js page by page for their text layer, annotation subtypes, image paint operators, attachments and JavaScript actions; images are decoded to pixels and scanned for transparency, near-black bands and EXIF, XMP, GPS or PNG text-chunk markers. It is for anyone about to release a document they have blacked out, and it returns one of three verdicts: high-risk evidence found, manual review required, or no obvious automated signal.
Open Redaction Proof Checker on AltFTool — it loads instantly in your browser.
Select the final redacted export you intend to share — PDF, PNG, JPEG or WebP up to 50 MB — rather than the editable source.
Press Check evidence to parse the text layer, annotation subtypes, attachments and scripts with PDF.js, or scan pixels for transparency, dark bands and EXIF.
Read the verdict, page counts and findings, then Download report writes <name>-redaction-check.json holding counts and rule ids, never the matched values.
A Redact annotation still present in the file is flagged high-risk, because the mark can exist while the content underneath was never removed.
Pattern hits are summarised as a match count and page numbers — the Aadhaar, PAN, email, phone or UPI string itself is not written into the report or the JSON download.
Every result ships with the limitations that apply: no OCR of image pixels, no recovery of overwritten pixels, no inspection inside password-protected content.
Often not. In a PDF, drawing a filled square or a Redact annotation over words can leave the original characters in the text layer, where anyone can select, copy or search them. Redaction is only permanent once the editor applies or burns in the marks and you save a fresh flattened copy — this checker looks for exactly that failure by extracting the text under the marks.
Six pattern families: Aadhaar-like 12-digit numbers, PAN-format identifiers (five letters, four digits, one letter), email addresses, Indian mobile numbers, UPI handles on the common bank suffixes, and credential words such as password, OTP, API key and private key. Any hit in extractable text raises the high-risk verdict, reported as a count and page list rather than the value itself.
No — a clean result is not proof of safe redaction, and the tool says so in its own output. It does not OCR pixels, reverse a blur, recover overwritten image data, or judge whether visible information is sensitive. Zoom to 100% and review every page yourself, ideally with a second reviewer, before you release anything consequential.
PDF, PNG, JPEG and WebP, and nothing is uploaded — parsing and pixel inspection run in the page. Large images are scaled so the longest edge is at most 1600 pixels before the pixel scan, which is why transparency and dark-band figures are reported against the inspected pixels rather than the original resolution.