The Only NotebookLM 2.0 Tutorial You'll Ever Need

0:00Notebook LM just had the biggest update in its history. Google rebuilt the whole engine behind it, so the tool that used to just summarize your PDFs can now run code, do its own research, and hand you finished work. So, I'll take you through everything that's new and exactly how to use it, because this is the version that turns Notebook LM from a study tool into

0:15something you'll get real work done in. When you open Notebook LM 2.0, the first thing you'll realize is that it looks just the same as the version you've been using before. If you've never really used it, Notebook LM was the tool you drop your sources into, your PDFs, articles, or YouTube links, and it would answer questions on them or turn them into an audio overview, all grounded in what you gave it. That part still works exactly the same, but the engine doing the work has been completely replaced, and that single change [music] is the whole reason people are calling this Notebook LM 2.0. Let's start with the

0:44model, because Notebook LM now runs on Gemini [music] 3.5, which is Google's newest and most capable model. So, the answers come back more accurate, and it holds together far better across a long messy notebook. [music] On top of that, there's a framework called antigravity, and this is the one that really changes things. [music] Antigravity is Google's agentic system, and agentic is just a word that means the AI can take real actions and use tools on its own instead

1:06of only writing text back at you. [music] It's the same technology Google built for its coding tools, where the AI can open things, run things, and work through a task step-by-step. Inside Notebook LM, that's what lets it go off and do the work itself, rather than sit there and answer questions about it. And what makes all of it real is that every notebook you create now comes with its own secure cloud computer. You never see it, and you never touch it, but it's a little sandboxed computer attached to your notebook that can write and run its own code with more than 100 built-in skills it can reach [music] for. In

1:34plain terms, that computer is what lets Notebook LM stop guessing and start doing, running real calculations and building real files from your sources. I'll give this notebook a question that needs proper work and not just a look up. Read through my sources and tell me where they disagree. And as you can see, you can follow every step it takes. It reads each source in turn, flags the points where two of them clash, and only then writes up what it found. That matters because I can check each step against the source it came from, which on research that really matters is a big deal. And none of this is a small bump

2:03because in Google's own testing, the new version beat the old one in roughly two out of every three of its main tasks with the biggest gains in handling long documents and finding good sources. So, the screen stayed the same, but the tool underneath went from something that reads and summarizes to something that can do the real work for you. The quickest way to feel that difference is the thing it can hand you at the end of a chat. And by the way, every prompt from this video is in a free playbook I put together, plus the workflow that chains all of it into finished work in a

2:30single chat. You can grab it in the description below. The clearest sign you're on the new version shows up the moment you ask the chat for a file. On the old notebook LM, the studio could put together a slide deck or a data table for you, but you were limited because the slides were just static images you couldn't edit, and the table would only ever export to Google Sheets, never to Excel. The new version changes that completely because you can now ask the chat directly for a PowerPoint, an Excel file, a PDF, a chart, or even structured files like CSV, and it builds the real working document for you. So,

2:59here's a notebook I've filled with research on launching a paid newsletter. To make a file, you just ask for it in plain words, so I'll type, "Build me an investor pitch deck from this research." A PowerPoint appears over on the right, and the first thing to know is that these new files are download only. So, instead of opening inside Notebook LM, it saves straight to your computer. And as you can see, it comes out looking

3:18like a real proper deck straight away. Once it's downloaded, you open it in PowerPoint. The deck it builds is structured sensibly, too, with a title slide, the problem, the numbers, and a closing ask. So, you're starting from a real first draft rather than a blank file. And that is the real change because you end up with a finished deck you can hand straight to someone. If something is off, you just go back to the chat, ask for the change in plain words, and it rebuilds the file with the fix. Now, for a different format, I'll go back to the chat and ask it to take that same research and lay out the cost

3:45and revenue projections as an Excel file. And it does the same thing here, a full spreadsheet with the projections already filled in. The same rule applies, it downloads rather than opening in the app, but once it's in Excel or Google Sheets, those are live cells you can edit and recalculate. For one more, I'll ask it for a one-page PDF summary of the whole plan, something I could send to someone who hasn't read any of the research. And as you can see, it comes back looking clean with the tables already laid into it. You can do the exact same thing for a Word document if you'd rather hand over a written

4:13report than a deck, and the structured formats like CSV are there for when you want to pull the clean data straight into another tool. So, in the time it takes to write a few short prompts, you've gone from a folder of research to a deck, a spreadsheet, and a report you could send out today. The files are the part you can see, but the bigger change is in how the notebook gets its sources in the first place. Normally, building a notebook means you go out and track down every source yourself, one at a time. On the new version, you can hand a lot of that job to the chat because it can read

4:38everything already in your notebook, look back over what you've discussed, and work out what you're still missing. The old search box on the left could never do this because it would go and search the web with no idea what was already in your notebook. So, back in that newsletter notebook, I'll type, "Based on what's here, what important sources are we still missing?" Instead of a generic search, it reads what I've already got first, then comes back with the specific gaps, things like total addressable market and industry growth, valuation multiples and exit strategies, and full operating expenses. And these are real gaps, the kind of thing I'd

5:07have missed on my own. From there, it uses Google Search to go and find solid sources that fill those gaps, and it lays them all out for me to look over before anything is added. You can also point it at a specific kind of gap, so if I ask it to find a couple of strong opposing views, it goes looking for the arguments against my idea rather than more of what I already believe. Nothing gets added without your approval because you choose which sources to keep, click import, and they drop into the panel on the left ready to use. So, you can start a notebook from a rough idea and let it

5:34build the source base alongside you instead of needing everything lined up before you begin. Once a notebook can find its own sources, run its own code, and build its own files, you can chain all three together into something the old version simply couldn't do. The next part changes how you use Notebook LM the most, and it comes down to whose data goes in. For this entire time, Notebook LM was built around other people's material, things like articles, PDFs, and videos. But now you can drop in your own raw files and have it work through them for you. This is where that secure cloud computer comes in, because instead

6:02of describing your numbers back to you, it can run real code over them, clean up messy or mismatched formatting, and do the actual maths. That formatting point sounds small, but it's usually what breaks this kind of work, because real data turns up in different shapes from different places, and the computer can line it all up before it runs a single

6:18number. So here's what that looks like. I'll upload two files here, one with my monthly ad spend, and one with my newsletter sign-ups, and I'll ask the chat to compare them and tell me which channels are converting. Because it can run code now, it doesn't just eyeball the figures, it works through them and gives me a clear breakdown of where the sign-ups are really coming from. And the breakdown it gives is clear and easy to follow. Then I ask to turn that into a chart, and it lays out a visual showing exactly what I'm spending against what each channel brings back. And the chart

6:43comes back clean and easy to read. What's useful is that it doesn't stop there. It talks me through what the chart is saying, so I can see which channel is losing money and which one is bringing in the most. That is Notebook LM doing the analysis for you, built entirely on your own numbers. And because of everything in the last two sections, I can take it one step further and ask it to write the whole thing up as a short PDF report with the chart built in, so the analysis and the finished document come out of the same chat. And as you can see, the report has

7:08the chart and the write-up sitting together, ready to send. It works just as well on something smaller and more personal. So in another notebook, I've been logging my newsletter's open rate week by week. I'll ask it to chart that open rate against the weeks I send a follow-up email, and it lines the two up together. And the pattern is obvious straight away, because the weeks I followed up are the weeks the open rate climbed. The reason that example matters is that it works on any numbers you keep, whether that's a training log,

7:31your monthly budget, or a year of sales. So Notebook LM has turned into a proper analysis tool, which reasons over your own data and shows it's working. Knowing what it can do is half of it, and the other half is a few habits that decide how good your results really are. The single biggest thing that changes your results is how you talk to it, because it can act on its own now, vague prompts get you vague work. So, the more specific you are about what you want and who it's for, the better the output gets. When I asked for that pitch deck, I didn't just say make a deck, I said

7:57make an investor pitch deck, and that one detail changes the structure, the tone, and what it chooses to put on each slide. It also pays to keep one notebook to one job. It's tempting to throw everything into a single notebook. The new version works best when each one is built around a single project, because everything it generates is grounded in what's in front of it, and a focused notebook gives it cleaner material to work from. Another habit worth building is letting it do the research before you start narrowing down. A lot of people still gather every source first out of habit, but you'll get a wider, more

8:24balanced base if you start with a rough idea, let it suggest the gaps, and only then pick what goes in. And the habit that ties all of it together is chaining the features instead of using them one at a time. The biggest difference comes when you let it find the sources, run the analysis, and build the document in one continuous chat, because each step feeds the next, and you end up with a finished piece of work rather than a pile of separate outputs. Get those habits down, and you're using the new notebook LM the way it was meant to be used. There's one thing left that

8:51decides how much you can trust everything it gives you. It comes down to thinking about a notebook a little differently. For years, it was the place you went to learn from other people's material, and it still does that brilliantly, but now it's just as much a place to make sense of your own. And if some of this isn't showing up in your notebook LM yet, that's because Google rolled it out to the AI Ultra plan first, with the cheaper Pro plan getting it soon. So, there's a good chance it's already there for you by the time you're watching this. In practice, you'll end up keeping two kinds of notebook, the

9:16ones full of other people's research, and the ones built around your own data, and the new version is happy doing both. That's the mindset to settle into, because once a notebook is grounded in your data, everything that comes out of it, in the studio or in the chat, is built only on what you put in, and that is exactly what made me nervous when I first saw this update, because the whole reason people trust notebook LM is that it answers from your sources and not from the open internet. With the chat now going off to find sources on its own, I expected that to weaken. But it

9:42holds up because it only ever brings in new sources when you approve them yourself. And when you're chatting, it still works only from what is in your notebook. So you keep the grounding that made Notebook LM worth trusting, and you get all the new abilities on top of that. The same Notebook LM you've always

9:55open can now do the real work for you. And that's the whole reason this update is such a big deal. And that ability to point Notebook LM at your own data and get real analysis back is the part I'd point you to first. Getting genuinely good at moves like that across every AI tool is the whole point of my school community, AI Fluency. So when you're ready to keep pace, go ahead and join with the link in the description. Thank you for watching, and I'll see you in

10:16the next one.