PDF Redaction Checker: Is the Text Really Gone?

What it checks. (1) Filled rectangles in the page\'s drawing instructions, classified as black, white or coloured boxes, with their position. (2) Text that can be selected but is not visible in the rendering: every character of every text item is tested against the pixels where it should appear, and a character whose area is uniformly one colour is reported as hidden. If a box lies on top it is named as the cause; otherwise the cause is "no box found", which also covers white text, tiny text and clipped text. (3) Redact annotations that were never applied: the content under them is still in the file. (4) Annotations such as opaque squares that cover text. (5) Optionally, words you type in: they are searched in the text of every page, in annotation text, in the document properties and XMP, in bookmarks and in attachment names, and every hit says where it is and whether it is visible.

Reading the result. A black box with selectable text under it is a likely failure. A black box with no selectable text under it is a good sign (the content was probably removed), but not proof, because text can also have been converted to shapes or pictures. White text on a black box is visible, so it is not flagged. Hidden text is shown shortened unless you tick the option to show it, and positions are given in PDF points from the bottom-left corner of the page, with an outline preview on the page.

What proper redaction needs. The content has to be removed from the file, not covered. Use a redaction feature that deletes the text and image content under the marked area and then flattens the result, and run the check again on the output. Also remember the document properties, bookmarks, attachments and comments, and the file name.

Performance. Pages are drawn at twice their size to look for hidden text, so large documents take a moment. The check handles up to 120 pages at a time.

PDF files up to 50 MB and 120 pages, with or without text. Scanned PDFs without a text layer have no selectable text to find, so only boxes and annotations are reported. Password-protected PDFs must be unlocked first.

1. Optionally type words that should be gone (names, numbers), separated by commas.
2. Drop the redacted PDF onto the box (up to 50 MB, 120 pages). The pages are analysed in your browser.
3. Read the summary and the findings by page. Look at the outlined previews to see where each finding is.
4. If anything is reported, redact again with a tool that removes content, then check the new file.

• Checking a PDF before publishing or sending it to a court, authority or journalist
• Finding out whether a redaction tool really removed the text
• Searching a document for a name or number that should no longer be there
• Finding redaction annotations that were marked but never applied
• Learning how common redaction mistakes look inside a PDF

The tool never certifies that a PDF is securely redacted; it checks for common failure patterns. It cannot see redactions that failed in other ways: white rectangles over scanned or image-only text, text converted to outlines or pictures that is still readable in the image, leaks through glyph positions or font subsets, content in hidden layers or other pages, earlier versions kept in the file, and unusual obfuscation or complex vector overlays. Character positions inside a text line are estimated by equal spacing, so edges of a box can be off by a character.

A report with no findings means these checks found nothing, not that the document is safe to publish. Treat any sensitive document as unsafe until you have looked at the redacted file yourself, tried selecting and searching the text, and checked the properties and attachments.