It's already over, the problem is the missing transparency. With an LLM you have no idea what influenced the answer, and there is no good way to show it to the user.