How to redact a PDF properly
True redaction removes information from the file. Anything less is a black box with the secret still underneath. Here is what actually leaks, what the federal rules require, and how to verify your work.
Updated September 5, 2026. Not legal advice; confirm the requirements of your court and any protective order.
True redaction versus black boxes
A PDF page is a set of drawing instructions: place this text here, draw this image there. Drawing a black rectangle on top adds one more instruction. The text underneath is still in the file, still selectable, still extractable, and still visible to anyone who deletes the rectangle or copies the page into another tool. That is how redaction failures make the news.
True redaction rewrites the page so the instructions that drew the sensitive content are gone: the text-showing operators are removed or cut, the affected image pixels are destroyed, and only then is a box drawn to show where something was. The output file contains no trace of the original content, in any layer.
What the federal rules require
Federal Rule of Civil Procedure 5.2 sets the baseline for filings in federal civil cases. Unless the court orders otherwise or an exception applies, a filing may include only:
- the last four digits of a Social Security number or taxpayer-identification number,
- the year of an individual's birth,
- a minor's initials, and
- the last four digits of a financial-account number.
The responsibility for redacting rests with the filing party; the clerk is not required to review filings for compliance. Parallel rules exist for criminal cases (Rule 49.1, which also limits home addresses to city and state), bankruptcy (Rule 9037) and appeals (Rule 25(a)(5), which applies the rule of the court below). State courts have their own rules, and many go further, covering driver's license numbers, medical information and home addresses. Protective orders in a case can require redaction of categories the rules do not mention. Read the rule, the local rule and the order, in that order. Each federal rule is quoted in full, with what to search for, under Federal redaction rules.
Where redacted information hides
A page that looks clean can still leak. These are the places to check, in rough order of how often they cause trouble.
- The text layer. The most common failure: a box drawn over live text.
- OCR text on scanned pages. The image may be blacked out while the invisible OCR text underneath still says everything.
- Earlier revisions. PDFs can be saved incrementally, appending changes to the end of the file while the original content stays inside. A redaction saved this way keeps the unredacted page in the same file.
- Metadata. The document's title, author, subject and keywords, and the XMP metadata stream, which often includes the original file name and editing history.
- Annotations and comments. Sticky notes, highlights and their pop-up text, including comments made during review.
- Bookmarks and links. Bookmark titles are free text and frequently contain names or subjects.
- Form fields. Field values persist even when a page is flattened badly, and hidden fields are easy to miss.
- Attachments and embedded files. A PDF can carry other files inside it.
- Hidden layers. Optional content groups that are switched off in one viewer may render in another.
- Cropped-out areas. Content outside the visible crop box is still in the file.
A method that holds up
- OCR first. Make scanned pages searchable before you redact, so search-and-redact can find every instance and so no OCR pass later resurrects the text.
- Find everything, then decide. Search by term and by pattern (Social Security numbers, account numbers, dates of birth) and review each hit. Accept or reject before applying anything.
- Apply true redaction. Remove the content, then draw the box. Never the other way round.
- Scrub the rest of the file. Metadata, annotations, bookmarks, attachments and hidden layers. Flatten forms if the values are sensitive.
- Save as a new file. A full rewrite, never an incremental save, so the earlier revision is gone.
- Verify the output with the steps below, then Bates-stamp the redacted file, not the original.
- Keep a redaction log listing each page, the reason and the term or area, so the production can be defended later.
How to verify a redaction
- Open the redacted file in a different viewer from the one you redacted in and use Select All on each redacted page. Nothing should be selectable inside the boxes.
- Search the file for each redacted term and for fragments of it. Zero hits everywhere, including bookmarks and comments.
- Extract all text from the file with a separate tool and search the extraction.
- Check the document properties for title, author, subject and keywords, and confirm there are no attachments.
- Confirm the file has a single revision by checking that it was saved as a new file, not an incremental update.
- For scanned pages, confirm the image itself is blanked, not just overlaid.
How ExhibitSet handles it
ExhibitSet's redaction removes the underlying text and image data using its PDF engine, then runs the verification above automatically: after applying, it re-reads the output and searches for every marked term, and it refuses to report success if any term is still found or any mark could not be checked. Every export is a full rewrite of the file, so earlier revisions and document metadata do not survive. Search-and-redact with pattern presets for Social Security, taxpayer and account numbers, a clean-out checklist for metadata, comments, bookmarks, attachments and forms, and a downloadable redaction log are all in early access today. Open the redaction tool.
Frequently asked questions
What is the difference between redacting and covering?
Covering draws a black box over the text while the text remains in the file and can be recovered by selecting it, extracting it or deleting the box. Redacting removes the text and image data from the file so nothing is left to recover.
What must be redacted in federal court filings?
Rule 5.2 of the Federal Rules of Civil Procedure requires filings to show only the last four digits of a Social Security or taxpayer-identification number, the year of an individual's birth, a minor's initials, and the last four digits of a financial-account number, unless an exception applies. Bankruptcy and criminal rules have parallel provisions.
Can I trust a redaction just because the page looks right?
No. Text can survive in the text layer, in annotations, in bookmarks, in metadata and in earlier revisions of the file. Verify by extracting the text of the output and searching for the redacted terms.
Does OCR affect redaction?
Yes. A scanned page has no text until OCR adds a text layer. If you redact the image but not the OCR layer, or run OCR after redacting a covered box, the text can reappear. Redact after OCR and verify the output.