A consultant exports a draft report to PDF to send to the client. The report itself is finished. Three margin comments from a colleague were still open, including one beginning "this isn't right, but we're not telling them that".
On most exports nothing would have happened. On this one, markup was displayed.
Export converts, it does not clean
Here is the thinking error. "Save as PDF" feels like a finishing move: the document becomes final, so whatever does not belong falls away.
What actually happens is a conversion. The application draws what it would print, and what it prints depends on how the document is being displayed at that moment and which options are set.
Which means the result varies. Two people exporting the same file can produce two different PDFs.
By application
Word. Comments and tracked changes live in the file until resolved. Whether they land in the export depends on the display setting and whether you export with markup. The reliable route does not run through export settings but through the source file: accept or remove everything before you export.
Excel. Hidden rows and columns are not printed and usually stay out of the PDF. The real risk is sharing the workbook itself, where hidden is one click from visible. Watch for hidden worksheets, and for formulas referencing an external file: that reference contains a path, and a path usually contains a username.
PowerPoint. Speaker notes travel as soon as you export a layout containing them. A plain slide export leaves them out. It goes wrong with the colleague who helpfully sends the notes version.
From a system. Exports from case or matter systems sometimes carry internal references, template names and user ids that never appear on the page.
And metadata on top
Every export writes fresh document properties: the author from the application settings, the producing software, a creation date, and on some routes the full path of the source file.
Detail that matters: this happens even if you carefully cleaned the metadata earlier. Clean, then re-export, and you need to clean again. See what stays in a PDF after you remove the text.
The check that catches everything
There are many export settings and they differ by version and operating system. Remembering which is where is not realistic.
What does work is one check afterwards, independent of how you exported:
- Open the PDF.
- Select all, copy, paste into an empty text editor.
- Read what appears. That is the text layer, and it is what a recipient or an AI tool extracts.
- Open document properties and look at author, title and path.
Two minutes, and it does not matter which application you used or which boxes were ticked.
This is also the only check that holds when uploading to an AI tool, because there nobody overlooks anything by accident: the tool reads the text layer exactly as it stands. See what happens when you upload a PDF to a chatbot.
