I’ve added product notes and internal documentation to a Claude Project, but when I ask a question, it isn’t always clear whether the answer came from those files or from the model’s general knowledge. That distinction matters when the documents contain newer details than the model might know.
Is there a good way to ask Claude to show which uploaded material informed an answer, or to check that it actually consulted the relevant file? I’m especially interested in how people verify Claude Project knowledge without treating a confident-sounding response as proof.