A PDF usually contains at least two layers: a text layer that holds the actual characters, and a drawing layer that decides what you see. Covering words with a black rectangle edits the drawing layer only. The characters are still there, still selectable, and still copyable.
The mistake that keeps appearing in the news
Most public redaction failures follow the same pattern: a document was annotated rather than rebuilt, and someone later selected the blacked-out region and pasted it into a text editor. If the text can be selected, it was never removed.
What real redaction requires
- Identify the text. On a scanned page there is no text layer at all, so the words have to be found with optical character recognition before anything can be removed.
- Review every match. Automatic detection has false positives and false negatives. A person should approve each candidate, and be able to mark regions the scanner missed.
- Rebuild the page. The safe output rasterises each page so the original glyphs no longer exist in the file.
- Verify the result. Check that no text layer survives, that the pages you kept still read correctly, and keep a record of what was applied.
How Vaulted PDF Redactor handles it
Vaulted PDF Redactor runs OCR locally with the document languages you choose, presents the candidates in a review list, accepts manual regions drawn on the page, and exports a fresh image-only PDF with a verification report. The original file is left untouched.
Two limits worth knowing: OCR accuracy depends on scan quality, and the export is image-only, so a signature or form that needs to stay interactive has to be handled from a flattened working copy. Review the export before it leaves your desk. Redaction is a process, not a button.