How to redact a PDF safely—not just cover the text

A page can look censored while still containing every hidden word. Many PDF editors let you place a black rectangle on top of text, but that rectangle may live in a separate annotation layer. A recipient can sometimes select the covered text, copy it, remove the annotation or extract the page's text with another program. Visual appearance is therefore not proof of redaction.

Real redaction changes the exported document so the protected content is no longer available underneath the visible mark. FeelPDF handles selected redaction areas by flattening the affected page to a new image. That deliberately trades away selectable text on those pages so the hidden material is not preserved as a text layer. Untouched pages remain copied as pages rather than being needlessly rasterized.

Duplicate the source document before making any changes. Work on the duplicate and keep the master somewhere access-controlled. Decide exactly what must be removed: names, account numbers, faces, signatures, barcodes, handwritten notes and information repeated in headers or footers. Search the whole document for each identifier because a name shown on page two may also appear in an index, bookmark or attachment.

Use consistent redaction boxes with a small safety margin around the sensitive material. Tight boxes can miss the edge of a character, while enormous boxes can remove useful context. Review at high zoom, especially on scans where text may be skewed. If a value appears inside a photograph or signature image, cover the pixels rather than assuming a text search found it.

The file that matters is the downloaded result, not the preview in the editor. Close the source, open the result in a different PDF viewer and try to select inside every redacted area. Search for the removed words and fragments of long numbers. Copy all page text into a plain-text editor and inspect the surrounding section. If a redacted page was flattened, it should no longer expose the original text layer.

Also inspect thumbnails, bookmarks, comments, form fields and file attachments. A person's name may be removed from the page but remain in a comment author field. Document properties can contain an author, company name, subject or original filename. Redaction and metadata cleaning are separate jobs, so do not assume one automatically completes the other.

Flattening is a strong way to prevent recovery of text that was under a box, but it changes the page. Search, accessibility and crisp vector output may be reduced on affected pages. A very low rendering resolution can make small print hard to read, while a very high resolution increases file size. Review the result at the zoom level and print size your recipient will use.

For court filings, public-record releases, regulated investigations or documents covered by an organizational retention policy, follow the required procedure and have another person review the result. A browser tool can perform the transformation; it cannot decide which facts the law or policy requires you to withhold.

Redaction material is sensitive by definition. Processing locally in the browser avoids uploading the unredacted source to a conversion service. You can open the browser network panel and confirm no document request is made while drawing boxes and exporting. Store the original and redacted copies with unmistakable names so the wrong version is not attached later.

Before sending, use a final two-person rule when possible: one person performs the redaction and another opens the exact outgoing attachment and repeats the search and selection tests. That catches both technical mistakes and the very ordinary mistake of emailing the unredacted file.