If I'm counting correctly, I've written 198 articles about NotebookLM to date. It's the first AI tool that managed to convince me AI has a genuine place in productivity, and it kept getting better with each update. 198 articles later, I finally got to ask the question I've always had: how does the team behind the tool actually use it?

While I've written a bunch about my own workflow and even covered a lot of tips from power users over on Reddit and X, I'd never gotten to hear it directly from someone who builds the tool for a living. That changed when I sat down with Michael Chen, a software engineer on the NotebookLM team who works on the Studio panel.

His notebook setup is split in two

Some are built to last, others are built to learn

NotebookLM is centered around, well, creating notebooks. You create a notebook, add sources to it, and then query them or use the Studio panel to turn them into something useful. What makes NotebookLM stand out from, say, ChatGPT or Claude Projects is that everything it generates is grounded entirely in the sources you provide. It doesn't pull from the wider web or its training data, and every response comes with citations pointing back to exactly where it got its information from.

The more I write and talk about NotebookLM, I've noticed that a lot of users don't really think about how they organize their notebooks. They just create a notebook whenever they need it, and let the homepage fill up and fill up. Michael's approach is more intentional than that, and it starts with a very simple distinction: some notebooks are meant to last, and others aren't.

On the personal side, Michael treats notebooks as disposable learning tools. He shared an example of creating a notebook for the chimapnzee civil war. He wanted to understand what was going on, so he spun up a notebook, used Fast Research to pull in relevant sources, got up to speed, and moved on. He did the same thing when he was looking up what shoes to wear for a half-marathon. These aren't notebooks he's going back to, and hence are essentially disposable. They exist to teach him something, and then they've served their purpose.

His work notebooks are a different story entirely. As an engineer on NotebookLM, he maintains a permanent notebook loaded with existing context about how the team builds features like Audio Overviews. That notebook follows him from project to project, so he never has to reload all that context from scratch. When a teammate proposes a new system or workflow, he drops it into that notebook and immediately generates a slide deck to get a visual overview. He then digs deeper through the chat with follow-up questions.

His notebook organization is surprisingly simple

I emoji-code everything, he just opens his favorites

While I usually only have good things to say about NotebookLM, something you'll find me constantly complaining about is the lack of organization the tool has. You can't create folders, tag your notebooks, and there isn't so much as a search bar to help you find what you're looking for. So, I asked Michael how he manages to keep things organized, and it turns out he doesn't really do anything special. Given that NotebookLM's homepage has an option to sort by most recently viewed, he simply opens his most important notebooks often enough that they stay in the first two rows.

This ties back to how his notebooks are split into two separate categories. The disposable ones naturally fall off the homepage as he stops using them, while his permanent work notebooks stay front and center because he's in them every day. Something I've been doing to keep my NotebookLM chaos organized is adding a prefix to the title of the notebook based on what I've created it for. For instance, if it's a notebook I created for studying, I'll add a "[College]" prefix to it. If it's a notebook I created for work, I'll add "[Work]" to the title. This way, I can hit CMD + F and search through my notebooks.

Given NotebookLM also lets you change the emoji of the notebook that's automatically generated, I also tweak the emoji to match the category. So, all my college notebooks get one emoji and all my work notebooks get another. The "emoji-coding" element doesn't really help with anything beyond giving me a quick visual cue when I'm scrolling through the homepage.

That said, for all the NotebookLM users wanting folders, don't you worry. At the end of our conversation, Michael flipped the script and asked me what NotebookLM feature I want most. I had 30 minutes with a NotebookLM engineer and the first thing I asked for was folders. You're welcome.

He doesn't necessarily start by uploading sources

Let NotebookLM find the sources for you

Given that you can't do anything in NotebookLM without creating a notebook first, I wanted to know what Michael's very first step looks like. For most of us (including me), the answer is obvious: you create a notebook, upload your sources, and start querying. Interestingly, Michael doesn't necessarily start by uploading sources. I mentioned this briefly above, but he shared that he turns to the Fast Research mode in NotebookLM to quickly pull in sources for him rather than hunting them down manually on Google.

If you aren't familiar with the Fast Research mode in NotebookLM, it's a research style within the Deep Research feature Google added to NotebookLM in November 2025. Deep Research has two research styles: Fast Research and Deep Research. The Fast Research mode rapidly scans the web and curates a set of relevant links you can add as sources to your notebook, while Deep Research goes further by generating a detailed summary alongside those curated links.

Michael defaults to Fast Research. His example was fairly simple too. He shared that he ran a half-marathon recently, and he used the feature to look up what shoes he should wear. Rather than opening Google Search, typing in a query, and filtering through results himself, he let NotebookLM do the fetching, and then had all his sources ready in one place to query, compare, or turn into a studio output.

When he shared this workflow, my follow-up question immediately was about the line between NotebookLM and Gemini. If you're using NotebookLM to search the web and pull in sources for you, how is that different from just using Gemini's Deep Research or any other AI chatbot?

Michael acknowledged the overlap, but for him the difference comes down to what happens after you get the sources. In Gemini, you get your answer and move on. In NotebookLM, those sources live in a notebook where you can generate Studio outputs, and keep coming back to them. You can populate your notebooks with more sources, and essentially create a living knowledge base that grows with you. That's the part a one-off Gemini search can't replicate.

The mobile app is his most underused recommendation

There's more to it than you think

I then asked Michael if there's a workflow or trick he uses in NotebookLM that most people don't know about, and his answer was something I didn't expect: the mobile app.

Frankly, this is something I need to get better at myself. I currently only use the NotebookLM mobile app to generate Audio Overviews when I'm getting ready in the morning. Michael thinks most people are in the same boat, and that there's a lot more the mobile app can do that people aren't taking advantage of.

His most-used Studio output isn't Audio Overviews

The most viral feature ain't his favorite

Audio Overviews is what started NotebookLM's entire boom. It's the feature that went viral, the one that put the tool on the map, and honestly, it's still the feature most people associate with NotebookLM. In fact, I asked Michael if the viral moment changed the team's roadmap, and he said it absolutely did and nudged them to focus on what users actually want. It led to new features like the 80-language Audio Overview expansion that launched last May, and I'd assume Video Overviews and Cinematic Video overviews were born out of the same momentum.

Despite Audio Overviews being the feature that made NotebookLM famous, it's not Michael's most-used Studio output. Instead, Slide Decks are. The feature, just as its name suggests, generates a full-fledged slide deck from your sources. When a teammate proposes a new system or workflow, his first move is to generate a slide deck just to get a visual lay of the land. From there, he'll dig deeper through the chat, ask follow-up questions, and sometimes generate another slide deck focused on comparing two different approaches side by side. I completely agree with Michael here — Slide Deck has quickly become one of my top NotebookLM features too.

I also asked him the Studio output he thinks is most underrated, and his pick was Cinematic Video Overviews. Unlike regular Video Overviews, which Michael mentioned people tend to dismiss as talking slideshows, Cinematic Video Overviews use fluid animations and rich visuals to walk you through complex topics in a way that's genuinely easier to follow. The feature is currently only available for paid users, but Michael thinks it really shines when you need to visualize how something works. As for regular Video Overviews, he thinks people are underestimating them too. He mentioned that the power isn't just that it's a slideshow with a voiceover. Instead, it's that NotebookLM stitches together a full narrative from your sources, and you get to consume it without doing any of the work yourself.

He has one tip, and he kept coming back to it

And it just takes a minute or two

Throughout our chat, something Michael really emphasized was the importance of spending some time customizing your NotebookLM experience. When I asked him if there's one setting or toggle he'd tell a power user to change from day one, he talked about the customization prompt that you'll find on most Studio outputs. For those unfamiliar, you'll find a pencil icon next to the generate button on Studio outputs like Slide Decks, Audio Overviews, and Video Overviews. Clicking it opens a prompt where you can tell NotebookLM exactly what you want like what angle to take, what to focus on, what narrative to follow. Most people skip it entirely and just hit the one-click generate button.

Michael says that's the single biggest thing he'd change about how people use the tool. He shared that if you just click on a Studio output (take Slide Decks for example) without customizing it, you're leaving it up to the model to decide what matters and what you want out of it might be completely different from how NotebookLM interprets your sources. But if you take the extra minute to write something as simple as "I want this slide deck to focus on the critique between option A and option B," the output becomes significantly more targeted and useful.

He genuinely believes customization features is the difference between NotebookLM being useful and NotebookLM being transformative. His example was a student who wants to understand Google's latest release about MCP. Without customization, NotebookLM gives you a general overview. But if you tell it "I'm a sophomore in college, I have a test coming up, and I need to understand this specific topic," the output shifts entirely to match your context. Without that extra minute, in Michael's words, there's no difference between NotebookLM and a general chatbot (and that includes Gemini).

The team is listening more than you think

Something I've loved about NotebookLM is how receptive the team behind is to user feedback. Michael shared that the team is very plugged into what users are saying on social media, and that they take it extremely seriously. He also recommended using the thumbs up and thumbs down buttons whenever something works well or doesn't work as expected!

That said, if you've ever posted a feature request, a complaint, or even just a workflow tip on X or Reddit, there's a good chance someone on the NotebookLM team has seen it.