Your guest is interesting — that's why you booked them. It's also why six other shows booked them first, and your listeners have heard those episodes. If your conversation covers the same origin story, the same three lessons and the same anecdote about the early days, you've produced a rerun with different microphones.

The prep that prevents this is simple and rarely done: know what they've already said. Not skim their biggest interview — know the record, so you can hear when an answer is a tape loop and steer somewhere new. That used to take a producer a week. It's now a half hour, which is exactly the gap between shows that get the rehearsed guest and shows that get the real one.

deep_search pulls the guest's appearance record — podcasts, interviews, panels, the small shows along with the big ones (small shows are where guests relax and say the unrehearsed things worth following up on).

run_youtube_tool compresses the record into prep: podcast-notes per appearance gives you each conversation's beats; key-quotes surfaces the lines they always deliver — which is your skip list — and the occasional line they said once and never repeated, which is your gold list. Thirty minutes of reading later you know their loops, their gaps and their threads worth pulling. Your first question can start where every other show stopped, and guests notice — the interview changes register when they realize you've actually done the reading.