繁中 →
visitors so far

The complete NotebookLM guide

All 9 tools in one go, broken down from running a one-person AI company

By Leo Huang · @kanisleo328 · compiled Apr 2026, the core workflow still holds

← Back to homeAbout me·Failure list·GitHub·Threads

How to read this guide

Full coverage: NotebookLM's (tested and written up Apr 2026; Google has shipped more tools since, these 9 core ones have not changed) 9 Studio tools(Audio Overview, Slides, Video Overview, Mind Maps, Reports, Flashcards, Quizzes, Infographic, Data Table), each with a screenshot, the steps, and the traps I actually hit. You also get how to create a notebook, how to pick sources, how to use YouTube videos as study material, how Deep Research runs, and how the new two-way Gemini integration works.

How it is written: follow-along style. I assume you clicked in today because you want to learn the tool, so there is no "what is AI" or "why NotebookLM" opener you already know. Every step has a screenshot, and every feature comes with when to use it and what went wrong for me. By the end you should be able to open NotebookLM and start.

Who this is for: office workers about to organize a pile of material with AI, one-person company owners, creators, students, researchers.

9
Studio tools (latest)
0
hallucination risk
Free
most features

If you already pay for Gemini, do not let this one go to waste.

Google AI Pro at $19.99/month = Gemini Advanced + NotebookLM Pro level + 5TB of storage, all in one subscription.
The free plan goes a long way (50 chats a day), but if you are already on Gemini, NotebookLM Pro access is included.

Jump to the tool guides ↓

First thing to get straight: NotebookLM and ChatGPT are not the same kind of thing

ChatGPT and Claude will talk about anything. Ask them a question and they answer from their training data. Flexible, but they can also invent an answer, known as hallucination.

NotebookLM works the other way. It answers only from the sources you upload, with nothing added. Give it three PDFs and it reads those three, and nothing else on the web. Every answer carries a citation marker, and one click jumps to the passage.

What does that buy you? It does not make things up. Ask it what a paper concludes and the answer is what the paper says, not something stitched together from the web.

This design idea has a name: source-grounded. It is the biggest difference between NotebookLM and other AI, and the reason it fits the case where you have a pile of material to understand, and does not fit the case where you want to toss ideas around.

Which one when?
· Answers from a specific set of material → NotebookLM (no hallucination, with citations)
· General knowledge, or asking AI to help you think up ideas → ChatGPT / Claude (flexible, but it can make things up)
· Analyzing very long documents (200 pages and up) → Claude (200K token context window, the largest)
· Writing code with AI → Claude Code (what I use every day)

1
Create your first notebook
30 seconds, a free account is enough

NotebookLM organizes everything into notebooks.

One notebook = one topic.
For example: competitor research in one, this month's YouTube learning in another.

The free plan gives you 100 notebooks with 50 sources each, which is plenty for most people.

Steps (click to open)
  1. Open notebooklm.google.com
  2. Sign in with your Google account (a Gmail account works)
  3. Click "Create new"
  4. Give it a name (for example "AI tool research")
  5. Done. You now have an empty notebook
2
Feed it good material and the AI gets smart
Garbage in, garbage out. Source quality decides output quality

How smart NotebookLM is depends entirely on what you feed it.

Give it one well-written article and it answers precise questions.
Give it a pile of messy notes and the answers come back messy.

Supported sources:

  • PDF — papers, reports, ebooks. Upload them or import from Google Drive
  • YouTube videos — paste the link and it pulls the captions (the video has to be public and have captions)
  • Web pages — paste the URL and it extracts the content
  • Google Docs / Slides / Sheets — import straight from Drive
  • Word / PowerPoint / plain text / CSV — all supported
  • Audio files (MP3, WAV) — transcribed to text automatically
  • Paste text — dropping raw text straight in works too
Limits worth knowing:
Each source holds up to 500,000 words (roughly 1000 pages of A4), with a 200MB file cap. The free plan allows 50 sources per notebook. Enough for 99% of what people do.
Steps (click to open)
  1. Open the notebook and click "Sources" on the left
  2. Click "+" or "Add source"
  3. Pick the source type (upload a file / YouTube / web page / Drive)
  4. Upload the file or paste the link
  5. Wait a few seconds while the AI processes it
  6. The source showing up in the left list means it worked. Click it to read the summary the AI wrote

Tip: you can add several sources at once. Keep one topic per notebook instead of dumping everything into one.

3
YouTube learning accelerator
A 3 hour video, the key points in 10 minutes

Watching passively vs asking the AI

Most people learn from YouTube by watching start to finish.

In a 60 minute video, maybe 10 minutes are actually about your problem.
You cannot tell which 10 minutes, so you watch the whole thing.

NotebookLM changes the flow:

Paste the video link
→ the AI pulls the captions
→ you ask "does this video talk about X?"
→ it tells you which part, quoting the line.

Passive intake becomes active questioning.
The learning speed is not comparable.

What you can do with a video

  • Automatic summary — once the video is in, the AI writes up the highlights, so you get the core points in 2 minutes
  • Interactive Q&A — ask anything about the video, and the answer marks which part of it came from
  • Turn it into a podcast — with Audio Overview, two AI hosts talk through the key points in conversation (Traditional Chinese supported)
  • Flashcards + quizzes — generated from the video automatically, so you can check you really understood it
Limits on YouTube sources:
The video has to be public (private or unlisted will not import). It needs captions, and YouTube's auto-generated ones are fine. Videos uploaded less than 72 hours ago may not import yet. Only the text of the captions is read, the picture is not analyzed.
Steps (click to open)
  1. In the notebook, click "Add source" → pick "YouTube"
  2. Paste the video URL (for example https://youtube.com/watch?v=...)
  3. Click "Insert" and wait for the AI to process it
  4. When it is done, ask in the chat area: "what are the three key points of this video?"
  5. For a podcast: click "Audio Overview" at the top → "Generate" → pick the language (Traditional Chinese)
  6. For flashcards: click "Flashcards" and the AI produces a set

Going further: add 5-10 videos on the same topic at once, then ask "what do these videos agree on? where do they contradict each other?" That one works very well.

4
Nine tools, nine outputs from the same material
April 2026 update: the Studio panel grew from 6 tools to 9

Same material, different ways to digest it

Upload one PDF and NotebookLM can help you understand it in nine different ways. You do not need all of them, pick by what you need. When I opened the Studio panel in April 2026 and counted, there were 9 tools:Audio Overview, Slides, Video Overview, Mind Maps, Reports, Flashcards, Quizzes, Infographic, Data Table. Plenty of Chinese-language guides still say six tools, which is out of date. Go by this page.

Audio Overview

AI podcast, two hosts talking through your key points

Turns your material into a 5-15 minute conversation show.
Two AI hosts discuss the key points in your sources, chat style.
Traditional Chinese supported.

Good for: listening on the commute, or getting through a long document fast. 3 a day on the free plan.

Mind Maps

Mind maps, see the shape of the knowledge

Pulls the core concepts out of your sources and draws them as a visual mind map. Good for seeing how ideas relate.

Reports

Reports, study guides, briefings, blog posts

Turns your sources into a structured report. Pick the format (study guide / briefing / blog post). Briefings export to PDF and PPTX.

Flashcards + Quizzes

Flashcards and quizzes, proof that you understood it

The AI writes questions from your sources to test how well you followed. Good for exam prep or checking what stuck.

Slides / Presentation (new in 2026)

Slides, turn your sources into a deck you can present

Picks the key points out of your sources and arranges them into a structured deck. Choose "detailed presentation" or "presentation slides", export PDF or PPTX, then keep editing in Keynote or Google Slides.

Good for: handing in a report, giving a presentation, running a workshop.

Video Overview

Video Overview, a Netflix-style short film

Turns your sources into a short video with narration, visuals and subtitles. Six visual styles to pick from.

Right now this one is open only to Google AI Ultra users ($249.99/month). Most people will not need it, and the other tools on the free plan already cover a lot.
Infographic (new in 2026)

Infographic, turn your sources into a visual

Generates an infographic from your sources. Good for social posts, illustrations in a report, or explaining one complex idea in a single image.

Added April 2026, and Chinese-language guides have barely mentioned it.

Data Table (new in 2026)

Data Table, unstructured text into a table

Finds comparable data in your sources and lays it out as a structured table. Good for competitor comparisons, feature matrices, price lists.

Added April 2026.


5
Deep Research, let the AI do the digging
The bottleneck in research is not finding material, it is organizing it

Deep Research in NotebookLM runs multi-step analysis on its own.

What you get back is a full research report, with a source behind every sentence.

Two modes:

  • Fast Research — a quick synthesis of your sources, results in minutes
  • Deep Research — a detailed investigation, searching the web or only your Google Drive

Available on the free plan with a usage cap, and the report can go straight back into the notebook.

6
Gemini + NotebookLM, syncing both ways
New feature, 2026/04/08

Why this integration matters

ChatGPT had Projects. Claude had Projects.
Gemini was the one without.

On April 8, 2026, Google's answer arrived:
NotebookLM goes straight into Gemini, and notebooks become Gemini's project system.

A notebook created in Gemini syncs to NotebookLM, and the other way around.

So mid-chat in Gemini you can point it at material you already organized,
with no re-uploading.

Where it stands: Open to Google AI Plus / Pro / Ultra subscribers on the web, with free accounts following soon. The details keep moving, so go by Google's official announcement.
7
How creators use it
Four ways I use it myself
Case 1

Competitor research

Drop 5-10 articles or videos from people in your field into one notebook and ask: "what three topics do they cover most? what angle have I not covered yet?"

Case 2

YouTube learning → content

After watching a few related videos, put them all into NotebookLM → generate the summary → make a "what I learned" podcast with Audio Overview → then write the takeaways up as a Threads post. One batch of material, three outputs.

Case 3

Gathering material for a long article

When writing a guide, put 5-10 related sources into NotebookLM first → ask "what do these agree on?" and "where do they conflict?" → use the answers as the skeleton of the draft.

Case 4

Customer records as a knowledge base

If you have a pile of customer questions (form replies, email threads), put them all into NotebookLM as an internal knowledge base. Next time someone asks something similar, ask the AI instead of digging through history.

8
Free or paid, which one do you need?
The free plan is enough for most people

Straight to the point: the free plan is already very complete.

Most of the nine Studio tools are free to use.
50 chats a day, 50 sources per notebook.

Paying mainly raises the ceilings (more sources, more chats)
and adds the advanced features (Cinematic Video).

Open the plan comparison (the official page wins)
ItemFreeAI Pro $19.99/month
Chats per day50500
Audio Overview3/day20/day
Sources per notebook50300
Notebooks100500
The nine Studio toolsMost availableMost available
Deep ResearchLimited runsFull
Cinematic VideoNot availableNot available (Ultra only)
Privacy: Google does not train its AI on what you upload, and your material is not made public. Google states this explicitly.
Back to the tool guides ↑

About the author

I'm Leo Huang, an indie developer in Taiwan, and right now I run a one-person AI company. No employees, no office. Code, design, community, copy, all of it is me plus a full set of AI tools.

I only tried NotebookLM recently, and it is genuinely good. I gathered how the people who use it well work, and every feature here comes with a screenshot from a run on my own Pro account, so you do not have to stitch a workflow together out of a pile of other guides. Just follow along.

Where Pro could not reach a feature (Cinematic Video Overview, which is Ultra only), I say plainly that I did not capture it, then cite a guide I read and trust, instead of pretending I ran it myself.

One-person AI company Heavy Claude Code user Works well with AI Co-founder of NaLi Match leoaido.com

Useful? Follow me

More AI tool guides and notes from running a one-person company

@kanisleo328

Useful? Cite this page and pass it on