How to Use NotebookLM: A Complete 2026 Walkthrough

janeiro 3, 2026

If you are still learning how to use NotebookLM, the tool you signed up for a year ago barely resembles the one running today. Google rebuilt it around Gemini 3.5, added code execution, and turned a simple document summarizer into something closer to a research assistant with its own cloud computer.

That matters because most tutorials online still describe the 2023 version: upload a PDF, ask a question, get a summary. That still works, but it is now the smallest part of what NotebookLM can do.

This guide walks through every core feature as it exists right now, then closes with one detailed, real-world workflow you can copy step by step. [Related: Best AI Tools for Productivity 2026]

What NotebookLM Actually Is in 2026

NotebookLM is a source-grounded AI research assistant built by Google, meaning it answers questions using only the documents you upload rather than pulling from the open internet by default. It uses Retrieval Augmented Generation to provide responses backed by web citations, which reduces hallucinations and lets it process complex documents through a large context window.

The June 2026 upgrade changed the underlying engine. NotebookLM now runs on Gemini 3.5 alongside a new framework called Antigravity, the same agentic infrastructure Google unveiled at its developer conference in May 2026. The most notable result: every notebook now comes with its own secure cloud computer, letting NotebookLM write and execute code natively for data analysis directly inside the platform.

Translation: you can now upload three years of spreadsheets and ask NotebookLM to actually calculate something, not just describe what it sees.

📥 Step 1: Add Your Sources

Getting started takes under a minute. Head to notebooklm.google.com, create a notebook, then add sources: PDFs, Google Docs, websites, YouTube links, and now images with OCR support and CSV files as well.

If you are missing material, the Discover feature can help. It suggests relevant web sources and can surface roughly ten curated results you choose whether to add to your notebook, so building a research base no longer means leaving the tool to search manually.

✅ Upload as many related documents as you have. NotebookLM’s context window now handles well over a million tokens on higher tiers, so thin research is rarely the bottleneck anymore.

💬 Step 2: Chat With Your Sources

Once sources are loaded, you can ask direct questions and get answers grounded in your material, with clickable citations pointing back to the exact source.

Sample prompts that work well:

“Summarize the disagreements between these three sources on [topic].”
“Pull every number related to [metric] and tell me which source it came from.”
“What questions would a skeptical reviewer ask about this argument?”

📊 Because responses cite their source, you can verify anything NotebookLM tells you in seconds instead of taking it on faith.

🎬 Step 3: Generate Studio Outputs

This is where how to use NotebookLM stops being a one-line answer, because the Studio panel now produces far more than a summary.

Audio Overview turns your sources into a podcast-style conversation between two AI hosts. Interactive mode lets you join the conversation live and redirect what the hosts cover.

Video Overview produces narrated slides pulling in quotes, diagrams, and numbers straight from your documents. You can steer it with prompts like “focus only on the cost analysis section” and choose a visual style, including whiteboard, watercolor, or a more classic slide look.

Data Tables and Mind Maps organize dense material into structured formats, useful for spotting patterns you would otherwise miss buried in text.

Flashcards and Quizzes are grounded strictly in what you uploaded, so a study deck built from your own textbook chapter tests you on that chapter, not a generic version of the topic.

⏱️ Video Overviews take the longest to generate of any Studio output, so start one early if you need it for a specific deadline.

🗂️ Step 4: Export and Share

NotebookLM now exports further than it used to. Presentations can leave as editable PPTX files with slide-level editing, and reports can move into Google Slides while keeping their formatting intact.

Audio Overviews download as WAV files, while Video Overviews support 80-plus languages for narration, which matters if you are producing material for a non-English audience.

📋 NotebookLM Plans Compared

Plan	💰 Price	📚 Sources Per Notebook	🎬 Studio Access	🧠 Best For
Free	$0	Limited	Basic Audio/Video Overviews	Casual, occasional use
Google AI Plus	$4.99/month	Expanded	Full Studio access	Students, individual researchers
Google AI Pro	Higher tier	100+ notebooks	Deep Research, higher daily limits	Freelancers, consultants
Google AI Ultra	Highest tier	Maximum	Priority access, largest quotas	Teams and power users

🧑‍💼 Real-World Use Case: Turning Client Research Into a Ready-to-Present Report

Here is a full workflow a freelance consultant could run this week.

The situation: A client hands you 12 PDFs (industry reports, competitor pricing sheets, and interview transcripts) and asks for a summary presentation by Friday.

1. Build the notebook fast. Upload all 12 PDFs at once. Use Discover to pull in two or three recent web sources that fill any obvious gaps, like a missing market-size figure.

2. Interrogate the material. Ask: “What are the three biggest disagreements between these competitor sources on pricing strategy?” Follow up with: “Which interview transcript supports or contradicts each of those points?”

3. Turn numbers into a Data Table. If pricing data is scattered across multiple PDFs, generate a Data Table to pull it into one structured view before you build anything client-facing.

4. Generate a Video Overview. Steer it with a specific prompt: “Focus on the competitive pricing gap and the interview quotes that back it up. Audience is a non-technical client, keep it under eight minutes.”

5. Export as an editable presentation. Pull the output into a PPTX file, adjust slide-level details to match your client’s branding, and you have a first-draft deck ready hours before the deadline instead of days.

This single workflow, research to structured data to a narrated presentation, is the clearest example of why NotebookLM in 2026 has moved well past “AI summarizer” territory.

FAQ

Is NotebookLM free to use?

Yes, the free tier covers casual use well, including basic Audio and Video Overviews. Paid tiers mainly raise your notebook count, sources per notebook, and daily generation limits rather than unlocking entirely different features.

Can NotebookLM analyze spreadsheets and data, not just text?

Yes. Since the Antigravity upgrade, every notebook runs on its own secure cloud computer that can write and execute code, which allows real data analysis on uploaded spreadsheets rather than a surface-level text summary.

What is the difference between Audio Overview and Video Overview?

Audio Overview creates a podcast-style discussion between two AI hosts, ideal for listening while multitasking. Video Overview adds narrated visual slides with diagrams and quotes pulled from your sources, better suited for presentations or complex concepts that benefit from a visual.

Does NotebookLM work in languages other than English?

Yes. Both Audio and Video Overviews support more than 80 languages, and non-English outputs now match the depth of the English versions rather than offering a shortened version.

How is NotebookLM different from asking ChatGPT the same question?

NotebookLM restricts its answers to your uploaded sources by default and shows citations back to the exact document, while general chat assistants often blend in outside knowledge. This makes NotebookLM better suited to research tasks where accuracy against a specific document set matters more than broad conversational range.

Final Takeaway

Knowing how to use NotebookLM in 2026 means treating it as a research and production tool, not a document summarizer. Upload broadly, question the material directly, and let Studio outputs do the heavy lifting for whatever format your audience actually needs.

Start with the client-report workflow above on your next project, and adjust the steering prompts until the output matches your exact deliverable. [Related: Best AI Tools for Productivity 2026]

Post Views: 377