Gemini Notebook: Stop Using It Like a Chatbot

Most people use Gemini Notebook wrong. Learn the source-first mindset and strategic steps to unlock its full research potential.

By Central
A guide to using Gemini Notebook effectively by curating sources and leveraging its unique features like deep research and code execution.
Highlights
  • Gemini Notebook only answers from the sources you give it, so source quality determines answer quality.
  • The sources panel on the left side is where the real work happens; chat and studio are downstream.
  • Since June 2025, each notebook has a secure cloud computer that can run code and produce charts and spreadsheets.

Most people fire up Gemini Notebook (formerly NotebookLM) and type a question. They get a vague answer, blame the tool, and never come back. The problem isn’t the notebook. It’s that they treat it like ChatGPT.

Here’s the hard truth: Gemini Notebook only answers from the sources you give it. Nothing else. No internet browsing (unless you enable deep research). No general knowledge injection. The quality of every answer is capped by the quality of your sources. That’s the foundational misunderstanding. Once you fix that, everything changes.

Gemini Notebook only answers from the sources you give it.

The Source-First Mindset

Think of the notebook as a research assistant who reads only the material you hand them. If you hand them random blog posts, you get random answers. If you hand them curated papers, transcripts, and your own notes, you get precise, cited answers.

The sources panel on the left side of the screen is where the real work happens. The chat and studio panels are just downstream effects. Most people skip source curation entirely — they dump in whatever and expect magic.

How to Actually Build a Research-Grade Notebook

Step 1: Name It Before You Do Anything

Click “Create new notebook” and rename it immediately. The name tells the notebook what subject it’s working on, and it keeps you organized across the 100 notebooks you can have on the free plan. Call it “Machine Learning Research” or “Client Case Studies Q1” — something specific.

Step 2: Fill the Sources Panel Strategically

Don’t add sources randomly. Ask yourself: what material would let someone answer my questions thoroughly? Then use these methods:

Fast research (search box inside the notebook): Type your topic and choose web or Google Drive. The notebook returns a list of possible sources. You approve each one before it imports. This method has been available since November 2025.

Deep research (for starting from zero): Turn on web search + deep research together. The notebook reads hundreds of sites, then builds a multi-page report with sources attached. You get 10 deep research queries per month on the free plan, 3 per day on Plus, 20 per day on Pro.

File types that actually work:

  • YouTube links (public videos with captions) — paste the URL and the transcript becomes a source
  • EPUB files (entire ebooks in one source)
  • Audio files (MP3, WAV)
  • Images (14 formats, up to 100 slides)
  • Google Sheets (up to 100,000 tokens)
  • Word docs, PDFs, CSVs, PowerPoint, plain text, markdown
  • Gemini chat exports (via the Gemini app’s “Add to notebook” feature)

Each source can hold up to 500,000 words. The size cap is the only real limitation.

Step 3: Use Checkboxes to Control Context

Every source has a checkbox. This is the single most underused feature. When you have 50 sources loaded and ask a question, the notebook tries to use all of them — which dilutes the answer. Untick everything you’re not asking about, then run the query again. The answer snaps into focus.

Sources stay in the notebook even when unchecked, so you can re-enable them later. After 5+ sources, the notebook automatically labels and categorizes them. You can rename those labels, move sources between them, or delete them.

Step 4: Configure Chat Settings Once Per Notebook

Click the “Configure chat” button in the chat panel. Two settings change everything:

  • Conversational style: Default (general research), Learning Guide (walks you through ideas), or Custom (type your own persona — “Answer like a PhD student”).
  • Response length: Default, longer, or shorter.

Set these once per notebook and forget them. The chat history saves automatically and stays private even inside shared notebooks.

Step 5: Turn Answers Into New Sources

Anything you save to a note (via the “Save to note” button) can be converted back into a source. Click “Convert to source” on a single note, or “Convert all notes to source” under the note section menu. This creates a feedback loop: every good answer makes the notebook smarter for the next question.

Common Mistake vs. Corrected Approach

What Most People Do What Works
Type a question without adding any sources Name the notebook, then import 3-10 curated sources first
Dump in 50 sources and ask vague questions Check only the sources relevant to your current query
Expect the notebook to browse the web by default Use the search box or deep research explicitly
Ignore file type limitations Use supported formats (YouTube, EPUB, PDF) — avoid unsupported ones
Never save chat outputs Pin valuable answers as notes, then convert them to sources
Leave chat settings at defaults Set style and length once per notebook

The Studio Panel: Where the Notebook Outputs Stuff

The right-side panel generates things a chatbot can’t. The two you’ll use most:

Data tables: Ask the notebook to pull numbers from your sources into a table. It labels columns automatically and exports to Google Sheets via the three-dot menu.

Add note: This turns back into a source. It’s the only studio output that feeds back into the notebook.

Other outputs include audio overviews (Deep Dive — two-host podcast; Brief — one host under 2 minutes; Critique — points out weak parts; Debate — argues both sides), video overviews (cinematic, explainer, short), reports (blog post, study guide, briefing document), mind maps, flashcards, quizzes, slide decks, and infographics.

Free plan limits: 3 audio overviews per day, 3 video overviews, 50 chats, 10 reports/flashcards/quizzes/mind maps. Everything resets on a rolling 24-hour basis.

One Thing Most Guides Miss

The notebook now runs code inside itself. Since June 2025, each notebook has a secure cloud computer. Ask for a chart, and it writes the code, runs it, and delivers a PNG, SVG, or Excel file. Ask for a budget spreadsheet, and you download a real .xlsx. This works on Pro and Ultra plans even without sources loaded — though desktop only.

That capability turns your research into actual deliverables. You’re not just reading summaries anymore. The notebook becomes a mini data analyst.

Questions answered
  • Why does Gemini Notebook give vague answers?Because it only answers from the sources you give it. If you dump random sources, you get random answers.
  • How do I build a research-grade notebook?Name it immediately, fill the sources panel strategically with curated material, and use checkboxes to control context.
  • What file types does Gemini Notebook support?YouTube links, EPUB files, audio files, images, Google Sheets, Word docs, PDFs, CSVs, PowerPoint, plain text, markdown, and Gemini chat exports.
  • What are the free plan limits?Free plan limits include 3 audio overviews per day, 3 video overviews, 50 chats, and 10 reports/flashcards/quizzes/mind maps, resetting every 24 hours.
  • Can Gemini Notebook run code?Yes, since June 2025, each notebook has a secure cloud computer that can write and run code to produce charts, spreadsheets, and other deliverables on Pro and Ultra plans.
Share This Article