It’s what models people have access to and how they use it. If you are not specific and overly broad it a search engine. If you move to the other end of the spectrum it can be incredibly powerful.
I tend to build projects with very limited sources for it to draw answers from and use complex instructions to guide how it interacts with the source material. It’s a tool. If you learn how to use it you can become extremely proficient/productive with it.
For example, stacks of PDFs may make sense to you and I as repositories of information. But is that the best format for AI to interact with the data? Probably not unless the actual formatting of the information on the page is relevant. I have been converting large groups of sources into databases containing page, line, and text data for groups of references for a given topic. Then you feed it the database and ask it to review and cross reference related lines of text. That serves as your “brain” for questions on a given topic.