BASALT · JOURNAL

Redacting a Word or Excel file that became a PDF

2026-08-07 · redact a word document converted to pdf

A converted file is not a clean file. When you redact a Word document converted to PDF you are working on an artifact that inherited properties, review markup, and hidden structure from a source you may never have opened. The right order is to clean the source while you have it, then treat the PDF as untrusted and inspect it anyway.

What the export carries across

Document properties travel. Word, Excel, and PowerPoint write the author, the last-modified-by name, the title, and the company from the Office installation into the PDF's document information dictionary and its XMP packet. The author comes from the Office user name rather than from the document, so a file drafted by one person and exported by another names the exporter.

The title field is the one that trips people. It comes from the document property, not the file name, and properties are inherited when a template or an earlier matter's file is reused, so a brief for one client can carry another client's name in /Title while every visible word is correct.

Comments can survive as PDF annotations, keeping the body text in /Contents and the reviewer's name in /T, depending on an export option that is easy to leave in whatever state the last person set it.

Tracked changes survive differently. If markup is displayed at export time, the exporter renders it: deleted text struck through, insertions underlined, change bars in the margin, and balloons in a side pane, all drawn onto the page as ordinary text that extracts perfectly. Accepting or rejecting all changes in the source before exporting is the only reliable fix, and hiding the display of markup is not the same thing.

The spreadsheet and slide cases

Excel adds problems Word does not have. Hidden rows and columns are hidden in the workbook, and whether they render in the PDF depends on the print area, on active filters, and on how the export was configured. Sometimes they are excluded and sometimes not, so you cannot assume either way and have to look at the output.

Hidden worksheets are the same problem in a larger size. A workbook with a calculation or backup sheet nobody has opened in a year is exported in full if the scope was the whole workbook rather than the active sheet. Even when a hidden sheet is excluded, a visible cell whose formula depends on it discloses the number.

Cells outside the print area, white text on a white background, and columns collapsed to zero width render as blank space or not at all, and if the export includes them the values sit in the content stream. Slides bring speaker notes. PowerPoint can export notes pages, and notes are where people write what they intend to say rather than show, so a notes export is a materially different document from the deck.

Clean the source first

Do this while you still have the original, because it removes the material rather than covering it. In Word, accept or reject all tracked changes, delete all comments, and clear the document properties for author, company, title, and keywords. Check for hidden text formatting, a character attribute that suppresses display, and for text in headers, footers, and text boxes outside the main story.

In Excel, unhide every row, column, and worksheet, look at what appears, and delete rather than re-hide what should not go out. Check defined names and the data model for references you do not want disclosed, and set the export scope deliberately. In PowerPoint, decide whether notes pages are in scope. Then export, and stop trusting the export.

How to redact a Word document converted to PDF

The PDF is now an independent artifact with its own carriers. Inspect it before marking:

exiftool -a -u -g1 converted.pdf
pdftotext -layout converted.pdf converted.txt

In the exiftool output, look for the author, the company, a title that is not the current matter, and an XMP history recording derivation from an earlier file. Struck-through deleted language reads as ordinary text in an extraction, so a document exported with markup shown is obvious within a few seconds, and comment text drawn onto the page appears there too.

Then look at what is not in the page text:

qpdf --qdf --object-streams=disable converted.pdf expanded.pdf
grep -a -i "confidential" expanded.pdf
pdfdetach -list converted.pdf

This surfaces annotation /Contents strings, form field values in AcroForm dictionaries, link targets, bookmark titles, and embedded file streams. An Office document containing an embedded object frequently produces a PDF with an embedded file.

Mark what remains and apply. Values recurring throughout the body are a search job, covered in redacting every instance of a name, while repeated headers are a region applied over a page range, covered in redacting the same area across many pages. Apply once, as a single generation, then verify with an independent parser: if the redaction was saved as an incremental update the pre-redaction page is still there, and grep -c "%%EOF" above one is the signal.

Why converting to PDF is not itself a cleaning step

There is a widespread belief that printing or exporting to PDF flattens a document into a picture of itself. It does not. The export writes text as text-showing operators with fonts and /ToUnicode maps so the result stays searchable, and it carries document properties across as a feature rather than a bug.

Converting also does not sever the lineage, since the XMP packet can record a document identifier and a derived-from entry linking the PDF to its source. What conversion reliably changes is which tools can see the content, which is why a cleaned source and an uninspected PDF produce the same accidental disclosures. See what a PDF still carries and how to check whether a PDF was really redacted.