NotebookLM: Agentic Chat, Code Execution, and Multi-Format Export

Google upgrades NotebookLM with Gemini 3.5, Antigravity, agentic web search, secure cloud code execution, and multi-format exports for AI Ultra users.

Google is beefing up NotebookLM, its document research assistant, with agentic capabilities in chat and more advanced reasoning, now powered by Gemini 3.5 and Antigravity. Each notebook receives a secure cloud computer where the tool can write and execute code for complex analyses, leveraging over a hundred software skills.

The chat gains autonomy: starting from a simple idea, NotebookLM searches for sources on the web, adds them to the notebook, and continues the synthesis, with the user retaining control over the selected sources, which are always attributed. Regarding deliverables, it produces a wide range of downloadable formats from the studio: PDF reports with charts and tables, Excel spreadsheets, PowerPoint presentations, visualizations (png, svg), structured data (csv, json), or images via Nano Banana, all editable afterward.

In Google's comparative evaluations against the previous version, the new iteration prevails in over 65% of cases on average, and up to 69.9% for long document analysis and 78.2% for web search and source discovery.

Global and limited to the web, the deployment first targets Google AI Ultra subscribers and Workspace customers with AI Ultra or AI Expanded access, with broader availability announced for later.