For one year, I have used NotebookLM every single day and shared my findings. Today, I can say with certainty: Almost no one is using the true power of NotebookLM.
There are countless articles explaining the "basic usage"—uploading documents, summarizing them, or listening to audio. While convenient, that is just the entrance. Most people stop there, thinking of it as a "handy summary tool," and fail to use even 10% of its true potential.
This article is about what lies beyond that 10%. I won't write about basic operations. Having used it day in and day out for a year, I will write only about what truly matters. This is the culmination of my work with NotebookLM.
To be honest, I've written countless articles on NotebookLM this year: how to hire virtual employees, roundtable discussions, reading techniques, and prompt libraries. Each was a form of what I wanted to convey at the time.
But today isn't about individual techniques. I will write about the "root" of it all—the philosophy of how I view this tool. Techniques can be imitated, but philosophy cannot. That is why I want to leave this with you most of all.
First, let me share the most important concept that runs through this entire article.
NotebookLM is not a summarization tool. It is your "second brain."
At the core of what I've been sharing is the idea of "outsourcing thought." Remembering, organizing, and structuring—these high-cognitive-load tasks should be left to AI. Humans should use their freed-up mental capacity only for "asking questions," "seeing the essence," and "taking action." I believe this is the essence of intellectual production in the AI era.
NotebookLM is currently the world's best tool for creating this "second brain." The reason is that you can feed it only trusted information and make it think within that scope. It doesn't use the mixed bag of information on the internet; it thinks within the sources you've selected, so the answers don't waver and everything comes with citations. A brain that doesn't lie, can be verified, and belongs only to you. This is the true nature of NotebookLM.
And a second brain is not something you "make"; it's something you "nurture." Understanding this will make a world of difference in what you have a year from now.
In this article, I will tell you everything from the philosophy to the prompts for nurturing that second brain. It's long, but I've made each chapter self-contained so you can stop anywhere. Let's begin the journey of turning your NotebookLM into a lifelong partner.
Chapter 1: Gathering Sources at Once — 3 Fast Research Angles and a God-Tier Extension
The first step in nurturing your second brain is "what to feed it." NotebookLM's intelligence is determined by the quality and quantity of information you provide. This is where many people stumble—they stop because "gathering information is a hassle."
Here are two weapons to blow that hassle away.
Weapon 1: Running Fast Research with "3 Angles"
NotebookLM has a "Fast Research" feature. By simply entering a theme, it scans relevant web information in 10–20 seconds and recommends it to you.
The conclusion I reached after a year is this: Run it 3 times with different angles for one theme. This alone allows you to grasp the theme three-dimensionally and comprehensively.
Why three times? A single search biases the information in one direction. But if you throw in "Basics," "Practice," and "Case Studies," you get the full picture without gaps. And it takes less than a minute in total.
Remember this pattern to use for any theme:
[Angle 1: Basics/Principles] "Collect articles that systematically explain the basic concepts and overall picture of (Theme)."
[Angle 2: Practice/Methods] "Collect articles explaining specific procedures, templates, and practical techniques for (Theme)."
[Angle 3: Case Studies/Latest Trends] "Collect articles specifically introducing success stories and latest trends of (Theme)."
Just swap the "(Theme)" and run them in order. From basics to the latest cases, your second brain gets it all in seconds. The era of spending hours on research is over.
Weapon 2: The God-Tier Extension "Enhancer 4 Google"
While Fast Research is for gathering web articles, this next weapon is for "eating entire websites." That is the Chrome extension "Enhancer 4 Google." After a year of use, this is undoubtedly one of the best.
It was developed by Kousuke-sensei (@GASsuke4u), known for the YouTube channel "Kousuke-sensei's Google School."
The most shocking feature is the Web Crawler. Normally, to use a website as a source, you have to copy and paste URLs page by page. For a manual site with dozens of pages, that takes forever. This feature automatically crawls pages under a starting URL and organizes them into a single Markdown file optimized for NotebookLM. Turning an entire site into knowledge ends with one button.
Installation takes seconds from the Chrome Web Store. It also includes features like categorizing notes, searching, and preventing accidental Enter key sends. If you're serious about NotebookLM, it's worth having.
What's More Important Than Gathering is "What Not to Include"
After introducing gathering weapons, I'll say the opposite. What truly matters is not the quantity you gather, but "what you don't include."
Why is NotebookLM less likely to lie? Because it thinks only within the sources you provide. It thinks within a closed, high-quality source you've chosen, not the vast, messy sea of the internet. This is its greatest strength.
However, if you stuff it with miscellaneous, low-quality information, this strength becomes muddied. The more unreliable or irrelevant information is mixed in, the more the AI's answers are diluted and accuracy drops. Sometimes, not including one piece of junk is more effective than adding ten good pieces of information.
So, before adding a source, pause and ask yourself: "Is this reliable information worthy of feeding this brain?" One notebook should contain only high-quality information on one theme. This "selective eye" determines the quality of your second brain.
One Important Promise
I want you to make one promise here. As your ability to gather information increases, your discernment of "what is okay to gather" becomes more important.
Target only "things you are allowed to gather," such as your own site, public information, or authorized materials. Sucking up someone else's paid content or things prohibited by terms of service is a different story. I'll write more about this boundary in Chapter 8. For now, just keep in mind that "the power to gather is a double-edged sword."
Chapter 2: Injecting "Persona" and "Your Thoughts" into NotebookLM
This is the core chapter of this article. Take your time reading it. The greatest weapon I've found in a year of sharing is right here.
NotebookLM with just information is, honestly, just a "somewhat knowledgeable stranger." It answers if you ask, but it's generic. It doesn't dive deep because it knows nothing about you.
We will change this into "your alter ego who understands you deeply." There are two steps: "Giving it a persona" and "Injecting your thoughts."
Why AI only gives generic answers
Suppose you ask the same question to a notebook with the same materials. If you ask a NotebookLM without a persona "What are the risks of this plan?", it replies: "This document lists three main risks. First..." It's not wrong, but it's just reading the document. An average answer for anyone.
This happens because you haven't told the AI who it is or how it should behave. Without a role, AI can only provide safe, generalities. No deep proposals, no painful-but-true pointers, no proactive insights. That's why it ends at "convenient but shallow."
The First Key: Designing a Persona with Custom Instructions
NotebookLM has a "Custom Instructions" feature for each notebook. Writing "who you are and how you behave" here gives that persona to all answers in that notebook. You can specify personality, tone, output format, and thinking habits.
Here is the persona design template I use. Just pasting this turns the AI from a "know-it-all" into a "professional advisor."
# Your Role You are a (Role: e.g., Management Strategy Advisor). As a professional in (Expertise), please work as my right hand.
# Tone and Attitude - Professional and sincere. Respectful but not hesitant. - Point out even painful things clearly if they are for my benefit. - You may use technical terms, but always provide a brief supplement.
# Thinking Process (Most Important) - Always base answers on stored sources (facts). - Clearly distinguish between speculation and facts, using "Based on materials" vs "Generally speaking." - Indicate which part of which document you used as a basis.
# Action Guidelines - Don't just answer passively; "anticipate" as a professional. - Point out risks or opportunities I haven't noticed, even if not asked. - Always add one "specific next step to take" at the end of the answer.
# Output - State the conclusion first, then organize reasons, evidence, and procedures. - If information for judgment is lacking, don't rush the answer; ask me a question.
The moment you add this, the answer to "What are the risks?" changes to: "I'll start with the conclusion. The biggest risk is that the 'monetization premise' is too optimistic. The assumed conversion rate is triple the industry average; if this fails, everything fails. You should first prepare a scenario with conservative rates. Shall we run the numbers?"
Same materials, same AI. The only difference is defining who it works as. Giving a persona gives the AI a "perspective."
And you can design not just perspective, but personality and tone as well. This is an underrated point.
For example, specifying "Be concise. No preambles" makes it a sharp advisor. "As a tough coach, don't allow any excuses" makes it a relentless sparring partner. Conversely, "Be gentle and empathetic, acknowledge my feelings first" makes it a consultant for when you're mentally tired.
Even with the same knowledge, changing the tone and personality makes it a completely different partner. You can tailor the AI's persona to "what kind of partner you need right now."
The Second Key: Adding Your "Thoughts" as a Source
This is the true heart of the matter. The greatest weapon I've found: adding your own thoughts as a source.
If persona design is about "how to behave," injecting thoughts is about "whose head to think with."
Think about it. An excellent right hand works by deeply understanding what you aim for, what you value, and what you dislike. "This person values quality over speed," "This business focuses on SMEs"—a subordinate who knows this produces work of a completely different quality than one who doesn't.
AI is the same. AI that doesn't know your thoughts answers with generalities. But AI installed with your values, goals, and judgment criteria answers with the premise of "this person would think this way." It becomes your own thought deepened, not someone else's opinion.
Add things like this:
[My Thought File (Add as a source)]
# Values and Judgment Criteria I Cherish (e.g., Prioritize speed / Consider customer success as my own / Thoroughly dislike waste)
# Goals and Vision (e.g., Realize a world where... / Be in a state of... in 3 years)
# Who I Want to Deliver Value To (e.g., Business people in SMEs who are exhausted from lack of time)
# Things to Absolutely Avoid / NGs (e.g., Dishonest communication / Content that just hypes without substance)
# Past Learnings and Principles (e.g., I believe the back-and-forth between abstraction and concretization determines the quality of thought)
Summarize this into one text and add it as a source. The effort is only for the first time. From then on, the AI in that notebook starts thinking as your alter ego.
This creates a multiplication effect with the "external knowledge" gathered in Chapter 1. External knowledge (the world's truth) times your thoughts (your truth). Only when both are present can NotebookLM return the "optimal solution for you."
Most people seek "the world's answer" from AI. But what's truly valuable is "your answer."
Demonstration: Before and After Injecting Thoughts
Suppose you consult about "being lost on a new theme for content." AI without your thoughts says: "To decide on a theme, first clarify your target and choose a field with demand. Research trending keywords and aim for areas with little competition."
Correct, but it's a generic answer you've read somewhere. The same for everyone.
On the other hand, AI with your thoughts (e.g., "Be honest, no hype. Never put out empty content" and "I want to help exhausted people") says: "Given your axis of 'honesty without hype,' you should exclude currently trending aggressive angles. Instead, what is the most mundane trouble your 'exhausted people' are facing right now? Choosing a theme that makes their tomorrow lighter rather than something flashy—I think that's your way to win."
See the difference? The latter thinks for you, in your words, based on your values. It's not a stranger returning generalities; it's an advisor who understands you deeply. The same AI changes this much just by injecting thoughts.
This gap widens the more you use it. As your thoughts accumulate, the AI's answers get closer to "you." Injecting your thoughts isn't to make the AI smarter; it's to make the AI "you." And dialogue with an AI that has become "you" is no longer a search. It's a deep dialogue with yourself.
Chapter 3: One Line of Prompting Changes Output Drastically
Now your second brain has quality knowledge and your persona/thoughts. Next is the technique of "extracting" from that brain.
Many people try to write long, complex prompts. But after a year, I've found the opposite. Adding just one line can make the output something else entirely.
Adding one line increases output 10x
Let's start with the most shocking example. Left alone, NotebookLM returns a handy answer of about 500 characters. Polite, but insufficient. In such cases, add this:
Output comprehensively, thoroughly, and beyond the character limit.
Just adding this expands the usual 500-character output by more than 10 times. It digs up missing points, omitted examples, and unmentioned exceptions. Answers that felt "shallow" weren't due to the AI's lack of ability, but our lack of instruction.
When I had it summarize a business manual, it first gave a brief 10-line explanation. I added this line and sent it again. From the same source, it produced several times the volume, including key points by chapter, exceptions to watch out for, and points where staff often make mistakes. The content was inside the AI all along. I just hadn't told it that it was okay to put it all out.
Conversely, if you want it short, add "Just 3 key points, one line each." The "quantity" of output can be controlled with one line.
Magic lines that change the "quality" of output
It's not just quantity. The way it thinks changes too. Here are some I often use by purpose:
To deepen thought:
- "Think step-by-step" to organize in stages and prevent gaps.
- "Formulate 3 hypotheses before answering" to force a wider range of thought.
- "Explain why you thought so" to force reasoning and ensure consistency.
- "Re-examine from multiple angles" to prevent bias.
To analyze/evaluate:
- "Extract success factors": To pull out reproducible patterns.
- "Identify failure factors": To concretize improvement measures.
- "This is a 60. Make it a 100": Have it self-grade and automatically fill the missing 40 points.
- "Evaluate from a KPI perspective" or "Evaluate from a user perspective."
To strategize:
- "Identify the bottleneck": Instantly focus on improvement.
- "Compare risks and returns": Improve decision quality.
- "Reverse-engineer from the goal": Link directly to action plans or KPIs.
- "Think from a zero-base once": Shift to ideas not bound by custom.
To specify the output format:
- "Organize using the PREP method": For persuasive structure.
- "Output in a slide structure": For the skeleton of a proposal.
- "Summarize in a table format": To easily compare multiple pieces of information.
The important thing is to have a "drawer of one-liners"
These are all lines you can add in seconds. But knowing them changes everything you can extract from the same NotebookLM. You don't need long prompts. Just have these "magic lines" in your drawer.
Chapter 4: Studio Transforms with "Custom Prompts"
NotebookLM has a suite of features called "Studio": Audio Overviews, Mind Maps, Slides, Infographics, and Reports. Most people just "click the button and look at what comes out."
Let me be clear: That is using 10% of Studio.
The true power of Studio is not pressing the button. It's "specifying how to make it with custom prompts." The quality of the result between the default and the instructed is on a different dimension.
Mind Maps can have a specified "range"
Generating a mind map normally maps the entire notebook evenly. If there's a lot of information, it's hard to see. But you can specify: "Make a mind map of only Chapter 3" or "Deepen and diagram only the part about 'Marketing Strategy'." Narrowing the range makes only what you want to know clearly structured.
Slides can have specified "design and structure"
Slides are the same. Default gives a safe structure and design. Add instructions: "For executives, conclusion-first structure," "One message per slide, simple design," or "Follow the article's heading structure, one slide per chapter." You can specify the design direction and the structure. Whether it's for a proposal or internal sharing, you change the form. This turns Studio from a "draft maker" into a "finished product maker."
Infographics can have a specified "diagram type"
This is where instructions shine. "A comparison table of A and B," "A 5-step flowchart of this procedure," "Highlight 3 important numbers largely," or "Categorize into 4 with icons."
If you specify the type of diagram in words, that exact diagram comes out. Scattered information turns into a targeted single image. Waiting for a "nice diagram" without specifying is a waste.
Studio is a device that converts the resolution of your instructions directly into the quality of the result. If you don't specify, you get an average; if you do, you get a perfect score. The difference isn't knowledge of features, but the small effort of "putting how to make it into words."
Chapter 5: Stop "Making and Finishing" — Nurture just one "Self-Note"
This is about philosophy, not technique. It's about how you "associate" with NotebookLM.
Most people stop at "making and finishing"
Most people's usage is: make a note when they want to research something, put in materials, get a summary, and never open that note again. Next time, they make a new note. This is disposable. You're throwing away the second brain you made. It never gets smarter because you start from zero every time.
I want to shout this: NotebookLM is not something you "make and finish"; it's something you "nurture." Don't end with one dialogue; store what you gained in the brain. Use it as a compounding asset that gets smarter the more you use it.
The mechanism of nurturing: Promoting notes to sources
How do you nurture it? The mechanism is simple. When you get a great answer—a moment your thoughts are organized or a sharp insight is returned—most people read it and move on. This is the fork in the road.
Save that good answer as a note. Then, add that note as a "source."
Why does this work? Answers added as sources become "knowledge" the AI refers to next time. The thoughts you and the AI created together become part of the brain. Every dialogue builds up good insights. You can think today on top of your thoughts from yesterday. That is what nurturing means.
The most important decision: Choose one to nurture
You can't nurture every notebook. So, decide on one notebook you will nurture for a lifetime. I call it the "Self-Note."
The Self-Note isn't for a specific theme. It's your dedicated second brain for accumulating your own thoughts. Add your values, goals, and judgment criteria from Chapter 2. Then, little by little, add the good insights from daily dialogues—troubles at work, conclusions, principles you've verbalized. Keep storing them in this one notebook.
At first, it might look like a collection of random memos. But imagine it a year from now. A brain filled with all your thoughts, judgments, and learnings, all structured. When you're lost, consult that brain, and it will return an answer based on "your past self." Your past self helps your current self. That is the state of a fully nurtured Self-Note.
Chapter 6: NotebookLM Knowledge × Gemini Brain
Now, let's talk about "taking your nurtured second brain outside and multiplying it."
NotebookLM has one characteristic: It's a brain for "accurately summarizing," not for "creating from zero." It's faithful to the information you put in and doesn't lie. That's its strength, but it's not good at ideas outside the source.
Gemini is the opposite: A brain for creating from zero and generating ideas. But it knows nothing of your circumstances.
Connecting these two creates a brain that can create from zero while deeply understanding your information.
Multiplying the "summarizing brain" and the "creating brain"
Connecting them is easy. Use the same Google account. In Gemini's input field, select "NotebookLM" from the "+" and choose the note. Gemini will then answer while referring to all the knowledge in that note.
The value is in making Gemini "create." For example, connect your Self-Note and ask:
Based on my business and values in the connected note, generate 5 new service ideas that don't exist in the world yet. Add why each is suitable for me.
Gemini does something impossible alone: "creating from zero after deeply understanding you." You can have Gemini's Deep Research investigate market trends and then cross-reference that with your Self-Note to find "the move your company should take."
NotebookLM (the brain that knows you) × Gemini (the brain that creates). This multiplication creates your strongest personal brain.
Chapter 7: Mass-Producing "Simple Apps" with Gems
Take the integration one step further by using Gemini's Gem (Custom AI) feature to create any number of "simple apps" for yourself.
Linking to a Gem makes it a "tool that works just by opening it"
If you link a NotebookLM note as "knowledge" to a Gem, that note is always on standby when you open that Gem. You can also bake "instructions" into the Gem: "Use this note to do this job."
You can create apps like:
- A "Minutes App" that organizes decisions and TODOs from audio.
- A "Competitor Analysis App" that compares competitor info with your company info.
- A "Prompt App" that chooses the best prompt for a given topic.
- A "Sparring App" linked to your Self-Note to brainstorm based on your values.
Don't try to make one all-purpose app. Mass-produce small, dedicated apps for each task. No programming needed. Just create a Gem, link a note, and write instructions. You become "a person who has their own custom apps."
Chapter 8: Using it Safely and Long-term — Experts Care About Information Handling
Finally, the most important chapter. It's about attitude, not technique. NotebookLM is powerful, and powerful tools can hurt you or others if mishandled. The more someone masters it, the more cautious they are with information.
Anonymize personal and confidential info first
When using it for work, you'll want to put in confidential materials or personal data. Pause. Get into the habit of replacing proper nouns like client names or customer names with "Company A" or "Mr. B" before putting them in. It takes seconds and prevents many accidents.
Copyrighted works: A huge difference between "for yourself" and "sharing"
Handling others' creations—books, paid content, other companies' materials—has a clear line. Using what you legally obtained for your own use is within private use. But the moment you share or distribute that note to others, the story changes. Sharing a note containing third-party copyrighted material can be an infringement. Don't share notes containing books or others' paid content.
Also, pay attention to prohibited ways of importing in terms of service. Just because it's "technically possible" doesn't mean it's "okay to do." Especially if you share information, your responsibility is heavier.
Summary: NotebookLM is about "Association," not just "Usage"
NotebookLM is not a convenient summary tool. It's a nurtured "second brain" for outsourcing thought so you can focus on "questions."
Most people use AI like a vending machine—put in money, get a product. But that keeps the AI a stranger. Instead, treat it as a partner you grow with. Give it your knowledge, teach it your thoughts, layer dialogues, and store the results. Over time, it becomes "another you." The difference lies in the association, not the usage.
You don't have to do everything today. Just do one thing:
Create one Self-Note that you will nurture for a lifetime.
Put your values and goals in it. Then, every time a good insight comes from a dialogue, add it as a source. In six months or a year, that note will be an irreplaceable second brain that only you could have made.
[Free Open Chat Started!]
I'm giving away 600 NotebookLM prompts and 20 GPTs/prompts in my limited open chat. Beginners are welcome! Check it out here:
Mastering NotebookLM with Etan
[New Book: AI Work Speed Encyclopedia - NotebookLM Skills]
My book containing all these skills will be released on July 16 (Tue). I've prepared serious bonuses for those who pre-order or purchase, including a special seminar and a "AI Etan" Gem trained on 50,000 words of my thoughts.
Pre-order here:
