PDF work for legal teams

Drawing a black rectangle over a name in a PDF annotator hides it from view and leaves the text in the file. Anyone can select it, copy it, or extract it with a script. This has produced real disclosure failures in real cases, repeatedly, and the pattern is always the same: the redaction looked right on screen.

Genuine redaction deletes the underlying text and replaces the region, so there is nothing left to recover. Verify it afterwards the way opposing counsel would: open the redacted file, select all, copy, and paste into a text editor. If a redacted name appears in the paste, the redaction failed.

Do the redaction before anything else in the pipeline. Redacting after merging means every intermediate copy you produced still contains the original text.

A scanned exhibit that has been through recognition carries an invisible text layer behind the image. It is what makes the bundle searchable, and it is also a second copy of every word on the page — including the words under your black rectangles, if the redaction only touched the image.

Check any scanned document for a text layer before treating a visual redaction as sufficient. If the text extracts, the redaction has to work on the text layer too.

A bundle is cited by page number, so the numbering must run continuously across documents that arrived separately and must be part of the document rather than a viewer setting. Merge first, then number, in that order — numbering the parts and then merging produces four documents that each start at one.

Where a header or footer needs to carry a case reference or a date alongside the number, add that at the same stage so the two are aligned and consistent across the whole bundle.

A completed PDF form keeps the answers in form fields, and some viewers — particularly phone mail apps and browser previews — render those fields blank. A recipient can open your filled form and see an empty one.

Flattening converts the field contents into ordinary page content, so what you see is what everyone sees. It also stops the answers being edited, which for anything being served or filed is usually the point.

Privileged and disclosable material processed through a web service is material that has been on somebody else's infrastructure. Whether that is acceptable is a question for your firm's policy, but it is worth knowing that it is a question at all — a file uploaded to a converter has left your control in a way a locally processed one has not.

Everything on this site runs in the browser, which means the document never reaches a server and there is no retention window to ask about. That is verifiable from the network panel rather than a claim you have to accept.

Recognition first, so there is text to find. Then the redaction, on the text as well as the image, and verified by extraction rather than by eye. Then removal of the pages that are not disclosable at all. Then merge into the bundle, then paginate, then flatten anything with form fields in it. Protection, if any, last.

The ordering is not arbitrary. Redacting after merging leaves the unredacted source copies in existence. Paginating before merging gives you four documents that each begin at page one. Flattening before filling makes the fields unfillable. And protecting before anything else means every subsequent step has to unprotect first.

Where this is done regularly, the sequence can be run as a single pass rather than as seven downloads and seven re-uploads of a document you would rather not have sitting in a downloads folder.