If you haven’t used NotebookLM much yourself but have heard of it in passing, chances are the feature you heard about was Audio Overviews. The two-AI-hosts-turn-your-PDF-into-a-podcast thing.

Depending on if you’ve always been a podcast person, it likely went one of two ways: you got obsessed and began turning everything into an Audio Overview, or tried it once, thought “that’s cool I guess,” and never really opened it again. Well, chances are you were simply not using it to its full potential. Or at the very least, not for what it’s genuinely good at.

You have a lot more control over Audio Overviews than you think

Audio Overviews are only as good as you make them

By now, I’m sure you’ve had enough people lecture you on how the quality of the prompt you send a LLM determines the quality of the output you get. The same thing applies here, and a lot of people don’t really realize it! Given how much I use and write about NotebookLM, I’ve seen how a lot of people use NotebookLM and specifically the Audio Overview feature, and it’s more or less the same way.

They drop a source in, hit the Audio Overview button in the Studio panel, and expect the podcast to somehow know exactly what they’re looking for. And then when it comes back sounding like two overly enthusiastic hosts giving you the Wikipedia version of your document, they write it off. What a lot of people miss is that you can actually guide the conversation before it’s even generated. Right next to the Audio Overview tile in the Studio panel, there’s a pencil icon that opens up customization options.

You can choose a format, adjust the length of the Audio Overview, and even add a custom prompt that tells the hosts exactly what to focus on. So, for example, if you’re preparing for a final and uploaded a bunch of lecture slides to your NotebookLM notebook, just hitting the Audio Overview button will get you a generic output that tries to touch on everything and ends up going deep on nothing. But if you hit the pencil icon and type something like “focus on the differences between X and Y, and explain them like I’m struggling to tell them apart,” you’ll get a completely different podcast that actually addresses exactly what you’re stuck on instead of recapping what you already know.

The same goes for the format options. Most people only ever use the default Deep Dive, but there’s also Brief if you just need a quick 1-2 minute rundown, Critique if you want the hosts to poke holes in your material, and Debate if you want to hear both sides of an argument. Each one produces a genuinely different output from the exact same sources!

Audio Overviews work as an excellent filter

Instead of a replacement for reading

Credit: Saikat Basu/MakeUseOf

I think something a lot of NotebookLM users don’t fully understand is that the Audio Overview feature isn’t really designed to give you an exhaustive breakdown of your source(s). It’s meant to give you a conversational, high-level overview. If you’re expecting the Audio Overview you generate to go over everything mentioned in your source document, you’re simply setting yourself up for disappointment. Despite the feature consistently being advertised as “Deep Dive(s),” it doesn’t really make a lot of sense to expect a 10-minute podcast to cover the same ground as, say, a 50-page document!

Rather than treating the feature as a substitute for actually reading something, I recommend using it as a way to filter out whether you should actually read something or not. For instance, if you have a 40-page report sitting in your inbox that you’ve been putting off, generate a quick Audio Overview of it, and you’ll know whether it’s worth your time or not within 10 minutes. If it is, you can then take it a step further and generate specific Audio Overviews focused on the sections that actually matter to you (which ties right back to the customization options I mentioned earlier).

The real value is in how they connect your sources

The more you feed it, the smarter the conversation gets

The real strength of NotebookLM, and the thing that sets it apart from asking ChatGPT or Claude to summarize a PDF, is that it’s excellent at synthesizing across multiple sources at once. The Audio Overview you get from one source versus, say, ten or twenty sources uploaded to the same notebook is a completely different experience. With one source, you’re essentially getting a podcast-style summary of that one document. Fine, but nothing you couldn’t get elsewhere. With multiple sources, the hosts start pulling threads between documents, surfacing contradictions you didn’t notice, and connecting ideas across sources in ways that would’ve taken you hours to piece together manually.

For example, say you’re researching a topic, and you’ve saved a few blog posts, a YouTube video transcript, a research paper, and your own notes. Upload all of them into one notebook, generate an Audio Overview, and suddenly you’ve got two hosts walking you through how these sources agree, where they diverge, and what the bigger picture looks like.

You can then take it a step further and use the customization prompt to steer the conversation. Something like “focus on where these sources disagree” or “highlight what my notes are missing compared to the research paper.” That kind of cross-source synthesis is genuinely hard to do on your own, and if you aren’t using Audio Overviews to help you do it, you’re missing out.

Audio Overviews aren’t just a gimmick

While I know a lot NotebookLM users use Audio Overviews in a similar way to what I described above, the majority still surprisingly view it as a magic summarizer that’ll do all the heavy lifting for them! It’s still one of the most underrated AI features out there, but only, only if you use it right.