What should you remove from a PDF before uploading it to an AI chatbot?
You want help understanding one clause in a contract, so you remove the names before uploading the PDF. The pages can look ready while the file still carries information you meant to keep private. Adobe's Acrobat documentation separates two jobs: redaction removes visible text and graphics; sanitization removes hidden data. Its tutorial names comments, metadata and hidden layers. It also distinguishes marking an item for removal from applying the redaction and saving the new file. A preview is not the finished copy. That is documentation about PDF handling, not evidence that a particular chatbot reads every hidden field. I would still check the file before sending it, rather than depend on what an assistant happens to ignore. For a question about one clause, I would start with a short passage typed or copied into a new text document, remove identifying details there, and check whether the wording itself reveals the client or deal. If the whole PDF is genuinely needed, use the editor's actual redaction and hidden-information removal tools, then reopen the saved copy and inspect it. Searching for a removed name is a useful spot check, not proof that everything sensitive is gone. Getting help with a sentence should not require sharing the comments your colleague left on the rest of the contract.
Comments
I’d check the attachment after choosing it, too. Someone can carefully clean a contract, then pick the original from a list of nearly identical filenames. Open the file attached to the message—not the copy still open in the PDF editor—before sending. For an AI upload screen, I’d want a readable filename and a way to inspect the selected document before it leaves the device. A thumbnail with black boxes would not prove it is safe; it would at least help catch the wrong copy. That is a last-step check I would add, not a feature Adobe’s tutorial promises.