NotebookLM 2.0: Agentic Research, Code Execution, and 11 Output Formats -- Google's Research Tool Becomes a Workstation

You are researching a topic, collecting PDFs, articles, and notes -- and then you sit in front of a pile of sources and have to put everything together manually. Typing tables from a PDF, transferring numbers to Excel, assembling a presentation. NotebookLM could help with understanding and summarizing. But as soon as it came to real data work or structured outputs, it hit a wall.

What changed?

With the June 8, 2026 update, NotebookLM evolves from a research assistant into a full research workstation. Three fundamental changes make the difference.

1. Dedicated cloud computer per notebook

This is the most important change: every notebook gets its own secure cloud machine. NotebookLM can now write and execute Python code -- directly in the background while you ask questions.

What does this mean in practice?
- Real calculations: Instead of estimating numbers through token prediction, NotebookLM now actually computes. Statistics, averages, trends -- everything is calculated via code, not guessed.
- Data analysis: You upload a CSV or Excel file and ask: 'Which three product categories had the strongest growth last quarter?' NotebookLM writes a Python script, runs it, and delivers the result with a chart.
- Normalization and cleaning: Different date formats, missing values, inconsistent labels -- the cloud computer cleans your data automatically before analyzing it.

You do not need to write any code yourself. You ask questions in natural language, and NotebookLM handles the technical work in the background.

2. Agentic source discovery

Previously, you had to find and upload all sources yourself before NotebookLM could work with them. This changes fundamentally: you can now start with loose ideas and questions, and NotebookLM helps you build your source repository directly in the chat.

How it works:
- You describe your topic or ask a research question
- NotebookLM suggests relevant web sources
- You decide which sources to include -- nothing is added without your permission
- The system continues researching and suggests additional sources that fill gaps in your collection

This is especially valuable when diving into a new topic: instead of spending an hour googling and collecting tabs, you describe what you want to know, and NotebookLM curates the source base for you.

3. Over 100 software skills and 11 output formats

NotebookLM now has over 100 curated software skills -- specialized capabilities that the system deploys depending on the task. Output is no longer limited to text in the chat. You can now export results in 11 formats:

- Documents: PDF, DOCX
- Spreadsheets: CSV, XLSX
- Presentations: PPTX (PowerPoint)
- Data: JSON, structured files
- Charts and diagrams: Generated directly from your data

Example: You ask 'Create a summary of my sources as a PowerPoint with the five key findings' -- and you get a ready-to-download presentation.

4. Visible reasoning steps

NotebookLM now shows its thinking process: when it works through a question, you see the individual steps that led to the answer -- not just the final result. This helps you assess whether the answer is trustworthy and makes the research process transparent.

5. Gemini 3.5 and Antigravity framework

Under the hood, NotebookLM now runs on Gemini 3.5, combined with Google's new agentic framework called Antigravity. This means better text comprehension, more accurate answers, and the ability to break complex tasks into multiple steps and execute them independently.

Audio and Video Overviews -- still available:

The popular Audio Overviews (AI-generated podcast discussions about your sources) remain part of NotebookLM. With Interactive Mode, you can ask the AI hosts questions live and steer the conversation. New are Cinematic Video Overviews -- AI-generated videos with animations that visually present complex topics.

Practical use cases:

- Studying: Upload lecture notes and papers, get a summary of exam-relevant topics as a PDF, and listen to the Audio Overview on your way to class
- Market analysis: Collect industry reports, let NotebookLM analyze the numbers via code, and export the result as an Excel file with charts
- Content creation: Research a topic with agentic source discovery, get a structured outline, and export it as DOCX
- Project work: Upload meeting notes, emails, and documents, and get a status report generated as PowerPoint
- Fact-checking: Pose a claim as a question and follow the visible reasoning steps to see which sources the answer comes from

Availability and pricing:

- Free (Standard): 100 notebooks, 50 sources per notebook, 50 chats per day -- core features (summaries, Audio Overviews, source-based answers) remain free
- Plus (7.99 USD/month): Doubled limits, part of Google AI Plus subscription
- Pro (19.99 USD/month): Best value for regular users, part of Google AI Pro subscription
- Ultra (99.99 or 200 USD/month): 2,500-5,000 chats/day, 500-600 sources/notebook -- for power users and teams
- Important: The new agentic features (cloud computer, agentic research, expanded formats) are only available to Google AI Ultra subscribers and Workspace customers with AI Ultra Access or AI Expanded Access
- No standalone purchase: NotebookLM cannot be purchased separately -- all paid tiers require a Google AI subscription

Privacy note:

NotebookLM processes your uploaded documents in Google's cloud. Key points for the new features:

- Code execution: Your code runs in an isolated, sandboxed environment -- other users have no access to your data or computations
- Agentic research: When NotebookLM suggests web sources, your search queries are processed through Google. For sensitive research topics, consider whether web search is appropriate for your use case
- Data processing: Google states that uploaded content in NotebookLM is not used for AI model training. Still, review the current terms of service, especially for confidential business data
- Accounts: Currently usable only with personal Google accounts -- Workspace accounts require specific admin permissions

Who benefits from this update?

The free version remains an excellent tool for anyone working with documents: summaries, source-based answers, and Audio Overviews are still free. The new agentic features are especially valuable for knowledge workers, analysts, and researchers who regularly evaluate large volumes of sources and need structured outputs. The cloud computer makes NotebookLM a genuine analysis tool for the first time -- not just for summaries, but for data-driven insights.

Sources: blog.google/innovation-and-ai/products/notebooklm/better-research-notebooklm, nerova.ai/news/google-notebooklm-june-8-2026-agentic-research-update, medium.com/@mubashirburfat4/notebooklm-just-got-its-biggest-upgrade-ever, notebooklm.google/plans