I’m considering using Perplexity to compare an internal planning document with public market information. Before uploading anything, I want to understand the boundary between analyzing a file and searching the web.

Suppose a document contains an unreleased product name, a customer name, and a distinctive sentence from our roadmap. If I ask for relevant competitors, could any of those details become part of a generated search query?

I’m asking about query disclosure, not only whether uploaded files are used for model training. Those seem like separate questions, and a setting about one wouldn’t necessarily explain the other.

Where can I verify what applies to a particular Perplexity plan and search mode? I’m looking for documentation covering uploaded document privacy, whether generated queries can be inspected, and which external services might receive them. Has anyone assessed this using synthetic confidential terms rather than real company data?