Overview
A document carries more than its text: who wrote and edited it, the firm's document management properties, the template's path on your network, hidden text, and every reviewer's comments and tracked changes. Metadata Clean downloads a copy without it, and tells you what it found. What it removes under each preset is on Metadata Removal.
Clean documents you drop in
- Click the Metadata Clean button (the broom) beside Selective Compare: between the Original and Modified columns on the home page, in the header of an open document, or in Quick Compare's toolbar.
- Choose a Preset:
- Standard removes what nobody sees on the page and keeps tracked changes, comments and authors' names.
- Strict also accepts tracked changes, removes comments and anonymizes authors.
- Change Tracked changes, Comments or Authors if this document needs something different from the preset.
- Drop Word, Excel, PowerPoint or PDF files, or a zip of them, into the panel, or click browse.
- Each file shows how many kinds of hidden data were removed. Click the arrow beside a file to see each kind, how many were found, and what was done, with any warnings to read before you send it.
- Click Download. Each file downloads under its own name.
Changing an option cleans every file in the panel again. Closing the panel deletes the files you dropped in and their cleaned copies.
Clean everything you download
Open any Download menu and turn on Clean metadata under Preferences. From then on, every document you download is cleaned with the standard preset first, including every file inside a zip. A document that cannot be cleaned does not download, so the original never goes out by mistake. The setting stays on in your browser until you turn it off.
In Claude or ChatGPT
With Version Story connected, ask for a clean copy: "Give me the NDA without its metadata", or "Send me a strict clean copy with the comments removed". The assistant downloads the cleaned file and tells you what was removed.
Read the warnings
Some things cannot be removed without changing the page, so they are left in place and listed as warnings: a document number in a footer, text under a black box in a PDF that is not really redacted, or white text over a picture. A signed PDF is cleaned, but its signature no longer verifies, so clean a document before it is signed.
Related
- Metadata Removal: what each preset removes
- Compare Privileged Documents in Claude or ChatGPT
- Files API: clean metadata