Transcription
Right now, the people getting the best results from Notebook LM aren't using some secret version of the tool. They just know which settings to change so they get faster research, more accurate answers, and responses that actually match how they think.
This matters whether you're using Notebook LM for learning, research, content creation, or just trying to make sense of the 47 PDFs sitting in your downloads folder. So, I spent the last few weeks testing every configuration option in Notebook LM and narrowed it down to five changes that genuinely matter. By the end of this video, you'll have a version of Notebook LM that is operating at its absolute maximum potential.
The most important update is the increased character limit on how you can customize Notebook LM's responses. Previously, you had about two sentences to tell Notebook LM how to respond. Now you have 10,000 characters, which is enough to build an entire instruction manual for how this AI should think. To access this, you press configure notebook on the top right, open the panel, select custom, and that's where you write your prompt.
Google actually shared example prompts when they announced this update and one of them turns notebook LM into a tutor that explains complex topics with simple language, real world analogies and checks along the way to make sure you're understanding. Which means instead of getting generic summaries, you can configure Notebook LM to teach you the way you actually learn. So let's say you're a student looking for a tutor support. I'll type, "You are a patient tutor helping me learn new concepts. Explain everything in simple language with real world analogies. After each explanation, give me one quick question to check if I understood. If I get it wrong, explain it in a different way." And now every single response from Notebook LM follows those rules.
But while that solves how the AI answers, it instantly brings us to the next problem, which is how much it can actually read. Notebook LM caps you at 50 sources on the free tier and 300 on plus. That sounds like a lot until you're writing a thesis or building a research project and you hit that wall. But there is a workaround that lets you bypass this limit entirely.
When you have a notebook full of sources, you ask Notebook LM to synthesize the key information into a comprehensive summary. Once you're happy with that summary, you save it as a note. And that note can become a source itself. So now you delete all your original sources and check only your newly created note. You've just compressed 30 sources into a single distilled source. And that means you've freed up source slots to add new material. You can repeat this over and over. Upload more sources, distill them into a note, clear those sources, add more, and distill again. I call this information alchemy because you're turning raw information into concentrated insight. You do lose direct citations to specific pages, but for personal research or learning, it's incredibly efficient.
I'll just go to one of my biggest notebooks and type something like "synthesize the key findings, arguments, and evidence from all selected sources into a comprehensive summary. Focus on the most important insights and how they connect to each other." And here's what we get. Once notebook LM gives me that synthesis, I click the pin icon to save it as a note. So now on the right, I have a distilled version of everything I just processed. [music] And from here, I hit the three dots and select convert all to source, which instantly moves that summaries right into my sources panel. This is how you build massive knowledge bases inside Notebook LM without ever hitting the limit.
But there's another feature that gets used completely wrong, and it's probably the most popular thing Notebook LM does. Audio overviews are the feature that made Notebook LM famous. You upload documents, click generate, and it creates a podcast style conversation that explains your material. But almost everyone uses this feature passively, and that's leaving a lot on the table.
When your audio overview is done processing, there's an interactive mode button you can press. Once you're in, the hosts start talking through your material. And at any point, you can hit join to jump into the conversation. You ask your question out loud, the hosts answer using your sources, and then they continue right where they left off. Think about what that means in practice. You're listening to a summary of a research paper. The hosts mention a term you don't understand. You hit join, ask, "What does that term mean?" And they explain it in the context of your specific documents. So, when you're jumping in to ask questions and getting answers grounded in your own sources, you're having a conversation with your research instead of just passively absorbing it.
But the audio player isn't the only place where hidden features are sitting in plain sight. If you look at the prompt input box right now, there are two drop downs that completely change how Notebook LM finds information for you. The first drop down is the research toggle, which lets you switch between fast research and deep research. Both of them search the web, but deep research goes significantly deeper. It spends more time digging, pulls in more sources, and can gather around 50 of them on its own before synthesizing everything into a response. So, instead of manually hunting for PDFs and articles to upload, you can give Notebook LM a topic and let it build the research base for you. Once it finishes, you'll see all the sources it found listed in your notebook, and you can remove any that don't fit or that hit payw walls. This is especially useful when you're starting a new project and you don't know where to start from.
The second dropown is right next to it and this one lets you switch between searching the web and searching your Google Drive. When you select Google Drive, Notebook LM can pull documents directly from your drive without you having to download and re-upload them. So you can type something like "find my invoices from the last month" and it'll surface those files for you to import, which is way faster than digging through drive folders yourself. You can start with nothing. Let Deep Research build a foundation from the web and then layer in your personal documents from Drive to add context that only you have. That's a research workflow that would normally take hours. And here it happens in minutes.
Everything I've just shown you lives inside your notebook. And you can share that entire notebook with anyone using a simple [music] link. When you share a notebook, the person receiving it gets access to all your sources and your configured AI. They don't have to upload anything themselves. They don't have to set up the customization. They just open the link and start using the expert system you already built. And this is where Notebook LM stops being a solo tool and becomes something you can build with other people. study groups, project teams, [music] client onboarding, courses you're building, literally any situation where multiple people need to learn from the same material. You can share with full access so people can add their own sources or chat only access if you want them to ask questions but not modify anything. Once you've built a notebook that works, you never have to build it again. You just share the link. To share, you click the share button in the top right corner and notebook LM generates a link just like sharing a Google doc.
Now, these settings only work as well as they do because of how I actually talk to the AI. And it all really comes down to prompting. [music] Doing that correctly is the difference between getting a generic answer and a perfect one. Google actually has a 6-hour course on this, and I sat through the whole thing, pulled out the lessons that actually matter, and compiled them into the video right here on the screen. So, click that, and I will see you there.