How to Use NotebookLM Better than 99% of People — Transcript

Notebook LM has grown in popularity, but I think it's still the most underrated AI tool out there. It's the number one tool I recommend to people in my everyday life when they ask me about AI. The capabilities have expanded massively over the past few months with new features and some powerful ways it can connect to Gemini that I don't think most people have realized yet. And it's still all free. There is a pro plan, but most people won't need it. I'll cover that at the end.

I'm logged in already, so I'll come down here and create a new notebook. And if you've never used this before, this will show you what it does along the way. You can upload sources across all sorts of file types. AI water usage is a topic that comes up a lot. So, I'm going to create a notebook around that. I watched this video from Hank Green covering it. He did a great job. He's awesome. So, I'll copy that link and then upload it. I'll add a couple of research papers I found

as PDFs. You can just drop them right in here. This website was very thorough. So, I'll drop in that URL. These sources are all solid, but I want a few more to round it out. One way to find more is with their discover sources option, which has two choices, fast research and deep research. I'll use fast research to start. AI water use. It goes out, searches across the web, and finds some sources. Now, it's important to go through and review these yourself to

make sure they're actually good sources. It does default to finding pretty good options, but it's certainly not perfect. Most of these are research papers, but right here I see one is from Reddit. I'll deselect that. Normally, I'd go to each site and vet the rest of these, but this is fine for the example. So, from there, I'll click import, and it brings them all in. And just like that, I've got a solid knowledge base to work with. Now, my notebook has all of that

information compiled, and it generated a summary for me right away. From here, there's a lot I can do. All my sources are on the left panel. The center panel is where I can chat with my sources. Let's use one of their default questions really quick. It reviews all the sources, sifts through the pages, then generates a comprehensive response, which, as you can see, has these in-text citations all throughout. It shows below a snippet of where that came from, or if

I click, it pulls up that exact spot in the source, so I can see it in full context. Notebook LM forces citations by default, so everything is grounded in real information. This massively reduces hallucinations, which is one of the main reasons I recommend it so strongly. But that is not all it does. The most powerful aspect is this studio panel on the right. That's where we can transform this information into all sorts of incredibly helpful formats. Podcasts, infographics, study guides, and the

quality is insane. I'll get to those in a little bit. First, I want to show you one other quick example of a completely different way to use this just to show that it's not limited to research and learning like most people think. Here's something I've been doing for a while with my YouTube channel. And I know that's not relevant to everyone, so I'll move through this quick, but it'll show you what's possible. In the analytics section, I can select all the parameters I want and export the current

view as a Google Sheet. This second column is the links to each video. And these are organized by view count. So, I'll just do the top 25 most viewed videos. And paste those all right in here. Click insert. And now I've uploaded each of my top 25 videos as a source. I also exported that whole spreadsheet as a PDF, so I can drop that in. Now I have a notebook with the transcripts for my top 25 videos, plus all my current analytics in one searchable place. I can ask it to spot

trends, identify what's working, or analyze my best performing content. But here's where it gets really powerful. I can also connect this notebook to the Gemini app and create custom AI assistants like a YouTube strategist or a script writing partner that have all this information already built in. I'll show you exactly how to do that later. Again, I know YouTube isn't relevant to everyone, but I'll go over other unique use cases at the end. I just want to

plant the seed right now of how flexible this tool actually is. But now back to my other notebook about AI water use. Something I do anytime I'm using this for research and deep diving into a topic is running a few prompts to check how good my sources are. I said Notebook LM is amazing for reducing hallucinations, but that doesn't help you if you have poor or biased sources. So, I start with this prompt. Looking only at the sources in this notebook, identify any areas where the sources

disagree with each other and any clear contradictions or conflicting claims. If I have a diverse range of sources and they all mostly agree, then I'm pretty comfortable moving forward with that information. If they don't agree, this helps pinpoint any areas they may either be wrong or just require nuance. I run that on every research notebook. I'll cover this again in a second, but I'll save that as a note and then delete the chat history and run prompt number two.

Based on these sources, what important questions or subtopics about AI water use are missing or barely covered? List the biggest gaps that would need to be filled to really understand this topic well. Do not invent details. Just describe what is missing. Sometimes your sources are incomplete and you can't know what angles are missing on a new topic. This prompt helps surface those blind spots. And I'm going to save that as a note and start the chat over with

this prompt. Are there any contrarian, alternative, or lesser-known viewpoints on AI water use that are likely not represented in these sources? Describe those possible viewpoints at a high level and suggest what kinds of sources I would need to look for to find them. This one's especially useful for controversial or evolving topics. The consensus view isn't always the correct one, and this helps identify alternative viewpoints you might be missing. For any

serious research project, I use one or more of these prompts before diving deep. Once I know my sources are solid, then I can start actually working with the information. From here, I can chat and ask questions like you would with any large language model, but with a crucial difference. Everything is grounded in information I can actually trust. It's extremely transparent about where every answer comes from, so I can fact check and get more context anytime I want. And as you're chatting, you can

check and uncheck sources to narrow the context however you'd like. Sometimes you may want to dig deeper into one specific paper or focus on multiple sources that are clustered around a specific topic. Then you can expand back to all of them whenever you want. The chatting itself is pretty straightforward at this point. Ask questions, get cited answers, but there are some settings that let you customize how the notebook responds. And these are worth adjusting based on what you're

trying to do. I'll come up to the configure notebook icon and first define your conversational goal, style, or role. Default is better for general research and brainstorming. Learning guide is optimized for educational content. And custom is where things get interesting. This is where you can really tailor it to your specific needs, like in the YouTube example I showed. This is where I'd tell it to act as my YouTube strategist. And this applies a lot to the examples I'll cover at the end. Then

under that, you also can choose your response length. I'm usually on default or shorter. I rarely use longer because in default mode, you can just ask for a specific response to be longer if you need it rather than making every single response lengthy. But overall, these settings are really helpful to adjust for each notebook. You se

leave and it will be there when you come back. But if you want to start fresh, you can come up here and click delete chat history. But before you do that, if you want to preserve your current conversation so you can come back to it later, click save to note. I usually save those initial three prompts I run as a note. And if you have a particular chat conversation that becomes really useful, like you've worked through a complex analysis or developed a framework, there's an option to convert

that into a source. then it becomes part of your notebook's knowledge base that you can reference in future chats. Now we'll get on to the really powerful stuff. This right panel, the studio panel has all these different formats to help you view, understand, and retain the information from your notebook. Audio overview, video overview, mind map reports flashcards quiz infographic, which is my favorite, slide deck, and data table. And just like with chatting, you can select and deselect

sources before generating any of these. So, the first two, audio overview and video overview, take a little longer to generate. So, I'll get those going first and come back to them. I'm generating a couple different options for each. I'll explain that later. And just a quick tip, any format that has a little edit icon means there are customization options you can adjust before generating. If you click anywhere else on the card, it just starts generating with the defaults. Let's move through

these others while we wait. Notebook LM is where I organize all my research into something I can trust. Then, I turn that research into content. And I use NADN to automate the tedious steps in between, like in our newsletter workflow here. I'm not going to walk through how to set this whole thing up right now, but I do have a full NAD crash course on this channel, and I talk about it a lot, so I assume most people watching are familiar. I self-host it using Hostinger. A lot of people think

self-hosting sounds intimidating, so they use NAD's cloud version instead, which starts at $20 a month. But self-hosting gives you more options, more control, and lower costs. The easiest way to skip the technical steps of setting up self-hosting is to use Hostinger's VPS. Unlike most providers, they've built a one-click installation template for NAD. So, you skip almost all the setup. I'll actually just walk through the whole setup to show you how easy this is. On Hostinger, go up to

services. Then, down under VPS, select the self-hosted N8 option. Pick a plan. The KVM2 plan is more than enough for everything I've needed to do and a lot more. and they actually have a great New Year's sale going on. So, right now it's $6.99 a month plus you get two months free when you go with the 24-month plan and they are the sponsor of today's video. So, make sure you add the coupon code futurepedia that gets an additional 10% off the sale price. After that,

standard billing and setup process. Then you'll see the dashboard and since I selected the NAD option, it's right up here. To open it, click manage app. Set up your NAD account. You can skip through this screen, but this one with the free license key. Click this button to send a key. Get that from your email. Go to your plan and click enter activation key and paste it in. And that's it. You are up and running with your own self-hosted version of NADN, one of the most useful and time-saving

AI platforms. I will have a link down in the description. And thank you to Hostinger for sponsoring this video. Mind map comes back really quickly. I'll open that up in full screen. It's got this nice little mind map. expand these out and collapse them back down, which is perfect for seeing how concepts connect and relate to each other. And if you click on one of these, it will read your sources and generate relevant information to that subtopic to read along as you go. That can be a great way

to visualize certain topics. Then when you click reports, you'll see several format options you can generate in create your own where you can craft reports your way by specifying structure, style, tone, and more. And there's briefing doc, study guide, blog post. And below are new ideas it generates based on your notebook. And they'll be different each time. And each one of these is further customizable. So I'll generate just a couple of these to demo. How about this risk analysis? And

one of the defaults like a blog post. Got this risk analysis. This super thorough report analyzing the environmental and social footprint of the AI value chain. And the other one was this blog post. It's got the whole blog post written out. And it looks like pretty good writing. This will vary massively depending on what type of topic you're using. Different types of reports will be helpful for different things. Next up are the study tools, flashcards and quiz. Both are

straightforward but really well executed. With flashcards, you can customize the number of cards and the level of difficulty and narrow it to specific topics or sources. The interface is really clean and intuitive. If you need to dive in further, you can click explain will automatically prompt for you and give you more details on whatever you're unsure about. And same things with the quiz feature. Customize the number of questions and level of difficulty. Narrow the topic if needed and you've

got a nice knowledge check. It has hints if you need help. If you get stuck on something, again, you can explain and jump right into the chat. The flashcards and quiz are great if you're actually trying to learn and retain this material long term. Next up is my favorite infographic. It uses Nano Banana Pro, which is incredible at this to create an infographic to visualize your data. And there's a ton of ways you can customize these and get creative with it. First, you can choose your

orientation. I'll stick with landscape and the level of detail. I'll generate one at each level. Then the customization box is where this really opens up. Now, for concise, I'll just let it do whatever it wants. For the standard, how about AI water usage, but style it like a comic book villain origin story. And for detailed, I'll let it just intake all the sources and see what it comes back with. But here's the simple version. This looks really nice. And I just read through this. It doesn't

have a single spelling error. All the text actually looks perfect. Although I don't have the short version, so it doesn't highlight the issues with including number three into the calculation. That gets much more complicated. Doesn't really fit into a concise infographic. But here's the comic book style villain's origin story. Rise of the hydrovore, AI's secret thirst for water. Only this is just amazing. And the text all does look completely perfect here. This is a

really nice style. It's even got a consistent character through it. Honestly, this is just incredible. Now, this is the detailed version. This is super aesthetic. Like, this just looks awesome. But this is where you start to run into the issue of errors in the text. It looks great at first glance, and a lot of this is right. But you'll find some little issues as you go through. Like right here, largest baders corsum electricity of 2 million households each. Like this is there's a

bunch of issues right here. Most of it looks good. A couple other spelling issues if you look close, so fair warning, you can definitely run into that. I typically stick to the standard or concise. Those tend to have little to no errors on them. And these are just so good at illustrating concepts visually. And slide deck also uses Nano Banana Pro. You can create a detailed deck, a comprehensive deck with full text and details, perfect for emailing or reading on its own or presenter slides, clean

visual slides with key talking points to support you while you speak. Start with a detailed deck and I'll use short as the length. And below I could add a highle outline or guide the audience style and focus. Generate this first one with nothing in there. And for presenter slides, I'll also use short. And we'll try something different like in a cyberpunk aesthetic with dark backgrounds, bright neon pink and electric blue highlights, glitch effects, make it feel like Bladeunner

meets data visualization. We'll see what that comes back with. So this is what the detailed deck looks like. These are all super aesthetic. It uses all the information from the notebook I have. Like it's kind of insane how good these can look. Like this page is just awesome to have the text over the top of these graphics and it fading into the background like this. This looks better than a lot of the dedicated slide deck tools. Let's take a look at this more

out there one we went with the ghost in the machine. Definitely looks like the vibe I asked for. And just a quick look through of this. This just nailed that vibe. So far, all of the text looks perfect on every single page. It's so insane to me it can make something like this. Like having this design where each of these icons actually represent what it says next to it. This actually having a diagram and highlighting different parts of a data center like that.

absolutely nailed it. This is basically as good as a slide deck can get with just a click from the information I uploaded. We've got one last one here, the data table. And like always, you can customize it tons of different ways. I'll just do one at the default and just let it come up with it on its own. I'll also try one compare the different sources of water usage across industries and see what that looks like. Does default have the environmental impact and operational statistics of AI models

and data centers broken down by the model names and different centers and how much water they consume in the region along with projections into the future. And it searched all those sources I had to extract all this information. This would have taken a huge amount of time to go through and find within those sources manually. Then the one I asked for broken down by industry. It's got a ton of data on here. This is the global 2027 projections for AI. It's like 4.2 to 6.6

billion. Sounds like a lot, but if you come down and look at agriculture, the US corn production, that's 75 trillion data centers use less than 1% of global freshwater withdrawals. And most people have a misconception on how serious the AI water issue is. But anyways, these video and audio overviews are done. And when you generate an audio overview, there's some different settings to choose from. A deep dive, a brief critique, or a debate. I started out with a deep dive, a lively conversation

between two hosts, unpacking and connecting topics in your sources. This is the one that sounds like a podcast, which went super viral when it first came out. And a cool part, which I'll show in a second, is there's an interactive mode where you can jump in yourself and ask questions or steer the conversation.

Welcome to the deep dive. Today, we're exploring a uh a crucial paradox,

a really big one, the incredible exponential growth of AI

and the absolutely massive environmental bill it's generating right now. We're sifting through your sources.

Water. The sheer scale of water consumption just to cool all these massive data centers.

And what are you talking about scale-wise?

Well, the estimates suggest that global AI demand could reach anywhere from um 4.1 to maybe 6.6 billion cubic meters by 2027.

That number sounds huge, but can you put that in perspective for us?

Get this. That could be nearly half of the United Kingdom's total water stressed. Oh, wait. Someone wants to join. Hey, go for it. Well, how do those numbers compare to other industries?

Yeah, that really helps put it in context

because context is everything when we talk about water use.

It's super easy to get misled by these huge numbers.

Absolutely. The truth is AI's water use is actually small compared to some other industries.

Here's where it gets really interesting.

Think about agriculture. One source compared AI to US corn production. Corn, that's a good one.

Corn alone requires around 20 trillion gallons of water per year for irrigation.

20 trillion. That's a massive difference.

Yes, that's nearly 80 times more water annually than all the world AI servers combined.

Wow. We're talking orders of magnitude larger.

And remember, 40% of that corn goes to ethanol for cars.

So, a lot of that water is powering our vehicles, not just food.

Exactly. Even something like watering US lawns uses trillions of gallons of municipal water every year.

That definitely helps frame the AI water use.

It does. It shows that while the AI numbers are big, the overall industrial and agricultural demands are far greater.

So, we're not saying AI's water use isn't a problem.

Not at all. It's a localized problem, and that's the key.

These podcasts are just awesome. It's a great way to absorb the data. Just a completely different format than all the other stuff in here. I like to generate one of those and listen to it when I'm in my car or doing the dishes or things like that. They'll jump into a bunch of different analogies throughout and frame it in ways that you would never see in like one of these papers. makes it much easier to digest and very helpful for me to retain it. So now for the video overview, you can create an explainer video, a structured comprehensive overview that connects the dots within

your sources or a brief, a bite-sized overview to help you quickly grasp core ideas from your sources. Then they have a bunch of different visual styles to choose from. Some of these are really, really cool. You can also create a completely custom style, which turns out great as well. This example I used heritage. And again, you can narrow the topic, source, or structure.

Today, we are diving into one of the biggest, most fascinating paradoxes of our time. The relationship between artificial intelligence and planet Earth. So, let's just get right to it. Here's the central question, right? On one hand, we hear all this amazing stuff about how AI could be the key to fighting climate change, but on the other, we know it has this massive physical resource-hungry footprint. So, which is it? Can it save us, or is it part of the problem? And you can really see that tension right here.

I think that's enough to give you a picture of how this works. I'll just kind of click through these as I'm talking. It's kind of like a combination of the slide deck and the audio overview allin one. It essentially made an entire slide deck that covers a full narrative and then created an engaging natural sounding voice over to explain it step by step. And it all syncs up with each one of these slides as it goes. And it's like a 7minute video overview. Incredible that it's able to do this. And I actually really like this style that I chose. Like some of these just

look really cool. Like this slide, it has some art from activist in Santiago, Chile. It brought that in and added this awesome design over the top. Like that's just insane. It's another one of those features that just blows my mind that this is possible. And I generated that explainer in the heritage style. I did also generate the brief in a whiteboard style. This is a much simpler design with less information on each slide, but it all looks great. Just depends on what

you need this for. Before we move on, there's one more feature I want to circle back to on the sources panel. So, earlier I showed you fast research for finding additional sources, but there's also a deep research option. This one takes a few minutes, but it's much more thorough. It uses an agentic model to search the internet on your topic and adapts its path based on what it finds and then eventually comes back with a full research report that you can add as

a source. Plus, it includes a list of all the sources it used, which is usually a lot. And you can pull those directly into your notebook as well. On the free plan, sometimes that will be more sources than the 50 you're allowed, and it will automatically select what it thinks are the best sources to keep you under that limit rather than just importing every single one. It's a great way to quickly build out a comprehensive knowledge base on a topic you're just getting into.

Now, here's something really powerful that most people have no idea about. You can access all your Notebook LM notebooks directly in the Gemini app. In any new chat, just click the plus button, notebook LM is at the bottom, and select a notebook. And now Gemini has all that context as grounding. Super useful when the Gemini interface is just more convenient, especially if you're revisiting something quickly. But here's where it gets even better. You can use Notebook LM as the knowledge

base for a gem. If you're not familiar, gems are like custom GPTs in chat GPT. They're specialized AI assistants you create by giving them instructions, context, and files. They let you handle repetitive tasks or access niche expertise instantly without having to start fresh every time. Like that YouTube strategist I mentioned earlier. This is where I'd actually build it. Not in Notebook LM itself, but as a gem connected to my Notebook LM notebook. If I create a new gem, give it a quick name

and description and add all the instructions for what I want it to do. I'll come down to the knowledge, click add files, and select notebook LM and add the YouTube notebook. Now, this gem has custom instructions on how to act, plus access to all my analytics and top performing scripts. This is honestly incredible. You can create endless specialized assistants, a work gem connected to a notebook with your company's data and SOPs, a fitness coach connected to your workout logs and

nutrition research, a competitor analysis assistant for your business. The Notebook LM connection makes building these knowledge bases so much easier than uploading files one by one. It's one of those integrations that just makes sense once you start using it. And I just asked it what video to make next. And this is actually pretty good. It combines two of my highest performing recent videos. You're not behind yet. How to build AI agents in 2026. I'm actually going to save that title and I

might make that exact video. Another thing worth mentioning is there's a mobile app for both iOS and Android. I'm typically on my computer, but the mobile app is great for a couple of things. The one is you can quickly save sources to a notebook on the go. Could snap a photo of a document, save a web page, that kind of thing. And number two is once you've generated an audio overview, it's way easier to just pull it up in the app and listen rather than downloading it from your computer and

transferring it over. I don't use the mobile app as my primary way to interact with Notebook LM, but for those two things, quick captures and listening to podcasts, it's really convenient. And if you're not at a computer all day like I am, you might find yourself using the app a lot more. Research and learning is what most people use Notebook LM for, but I don't think it crosses most people's minds how flexible this tool actually is. So, I'm going to run through some unique use

cases to get you thinking. And these won't all apply to you, so I'll move quickly. I already covered some YouTube use cases earlier, but here are a few others. a tech manual database or homeowner's binder. For every tech product or appliance you own, like a refrigerator, water heater, microphone, your car. Find the owner's manual online. Most are really easy PDFs to find. Then upload them all into one notebook. Now, anytime something breaks, you have every manual searchable in your

pocket. Personal journal analysis. If you journal, upload all your entries. You could just have a searchable archive, or you could ask it to identify blind spots, recurring themes, patterns, what you complain about the most. You could even connect it to a gem and give it instructions to act as a cognitive behavioral therapist. Or use the infographic feature to visualize your life's journey as a hero's journey in the style of Lord of the Rings. Whatever works for you, a fitness and nutrition

coach. Upload your workout logs, meal plans, and any fitness advice you follow. Ask it to analyze your progress. Suggest routines or create grocery lists based on your goals. You have meeting notes archive all your work meetings in one searchable place. Track progress. Identify recurring pain points. See what keeps coming up across conversations. Debate prep. Upload research on both sides of a topic. Then use the debate feature in audio overview to hear arguments played out. Competitive

research synthesis. Upload competitor white papers, reports, and website content. Ask Notebook LM to compare pricing strategies. summarize innovations or create comparison tables for team presentations. You a book club companion. It can handle an entire PDF of a book. Upload it, discuss themes, and it'll pull specific quotes and relevant passages as you go. Write the book that wasn't written. So, here's a creative one. Upload every interview from someone. Like, an example would be

Mr. Beast, every interview he's ever done, then ask it to write a book from his perspective about scaling a media company. Anywhere you have a collection of information that you want to actually use and interact with, Notebook LM can turn it into something practical. Let me quickly cover what comes with the Pro plan. Although for most people, there's no need to ever upgrade. On the free plan, you can have up to 100 notebooks with 50 sources each. The paid plan doubles that, but honestly, most

people won't come close to those limits, and sources can be huge, so you could even merge multiple documents into one if you need to. Where you'll feel the free plan limits more is with the studio features, audio and video overviews and nano banana pro generations for infographics and slide decks. You get a limited number of these each day on the free plan, but significantly more on pro. I had the free plan for a long time. The main reason I have the pro plan is because it's bundled with

Gemini. So I upgraded my Gemini plan to pro to access, you know, Gemini 3, Nano Banana Pro, VO, and some of those other premium features. And there's probably a lot of people out there that have done that as well. If you have Gemini Pro, you automatically get the higher notebook LM generation limits included. And Pro also adds the ability to share chat only notebooks with other people. They can see all your sources and chat with them, but they can't make any changes to the notebook itself. So

that's useful if you're collaborating on research or want to share a knowledge base with your team. It has an analytics feature to track that all as well. For most individual users, just doing research and learning, the free plan is more than enough. But if you're using the studio features heavily or already have Gemini Pro, it makes sense. Notebook LM is legitimately one of my favorite AI tools, and I only find myself using it more and more. So, I hope this helps you actually master it

instead of just scratching the surface. And if you want to go much deeper into learning all aspects of AI, we have a full course platform on Futuredia with over a thousand lessons across over 30 AI courses. You'll find full learning paths on everything from chatabt to video generation to coding with AI and everything in between. We already have a full course on notebook LM in there. It's all included in one subscription. You can get a 7-day free trial using the

link in the description. Or if you want to go deeper into more free Google tools, check out this video next.