How the checks work
Plain English for each check in the redaction verifier. The tool only reads your PDF in this browser tab. Your file is not transmitted. A clean result means these checks found nothing — not that the file is certified safe. A flagged result means text is underneath something opaque — check it yourself — not that someone deliberately concealed it.
4.1 Text under black marks / covering images
Many “redactions” draw a black rectangle on top of text and leave the letters in the page content. Others paste an opaque image (for example a black PNG) over the words instead of a vector fill. The checker walks the page drawing instructions for filled shapes and for images, measures each text glyph’s box, and asks whether those boxes sit under a dark fill or an opaque covering image. If they do, the characters are still in the file and can be copied or extracted. A flagged result here means text is underneath something opaque — check it yourself — not that someone deliberately concealed it. When an image sits over glyphs but its opacity cannot be proven, the check reports POSSIBLE (inconclusive) — never a clean pass for that case.
4.2 Invisible text
PDF text can be drawn in invisible render mode (operator Tr 3), or in a colour that matches the background (for example white on white). It does not show on screen, but it is still in the file. The checker looks for both patterns.
4.3 Hidden OCR layer on scans
A scanned page is often a picture with a text layer added behind it for searching. If someone blacks out the picture but leaves the OCR text, the words remain. The checker looks for image pages that still carry an invisible or camouflaged text layer.
4.4 Annotation-only redaction
Some tools drop square, highlight, or “redact” annotations on top of the page and never remove the underlying content. Annotations are markup. The checker reports those markup types when present.
4.5 Earlier versions
PDFs can be saved incrementally. Extra %%EOF markers and a /Prev pointer in the
cross-reference table mean an older revision may still be sitting inside the same file. The checker
counts those markers and looks for /Prev.
4.6 Metadata
The document information dictionary and XMP packet can hold author, title, subject, keywords, creator, producer, and dates. The checker lists what it finds and flags values that look like a person’s name, a file path, or a client label.
4.7 Embedded files
PDFs can carry attachments. The checker looks for embedded-file structures and file-attachment annotations and reports their names when it can.
4.8 Form fields
Filled form fields keep their values even when the page looks blank. The checker reports field names and values it can read from widgets and AcroForm data.
4.9 Embedded JavaScript
Some PDFs include JavaScript actions. The checker reports that JavaScript is present. It does not run it.
4.10 Bookmarks / page labels
Outline bookmarks and custom page labels sometimes still say things like a party name or a file title. The checker lists bookmark titles and non-default page labels when they appear.
What the verdicts mean
Recoverable text found. This document is not safely redacted. — Checks 4.1, 4.2, or 4.3 failed (including text under opaque covering images).
Other risks found. No recoverable text under the marks, but something else can leak. — One of 4.4–4.10 failed, or check 4.1 reported POSSIBLE (image cover with unproven opacity).
No problems found by these checks. — Everything that ran passed. Not a safety certification.
This is a scanned document. No text layer at all — these checks cannot tell you much. — Image-only pages with no extractable text.