When working with documents, the citations that NotebookLM provide increase the confidence in answers. This allows to use NotebookLM for team-level knowledge bases.
U can do that with chatgpt which preceded notebooklm
Podcast generation is quite interesting as a concept, but if you listen to more than a few hours of podcasts a week, you'll probably find it uncomfortable to hear anything coming out of NotebookLLM -- too unnatural, the pace is off, and the thing feels more like two robots taking turns to read a script with exaggerated reactions than two humans chatting with pauses and interruptions. At least I stopped after 3 minutes of that.
I was putting together an essay/podcast about a news event in another country, and I was able to collect multiple YouTube videos, search and translate their transcripts, and ask questions like 'is this the video where someone talks about _?'