What NotebookLM’s Agentic Research Upgrade Really Is
NotebookLM’s agentic research upgrade is a new AI research automation layer that uses Gemini 3.5 and cloud code execution to plan, run, and refine complex research projects with less manual prompting while keeping users in control of sources and outputs. Built over three years from a Google Labs experiment into a full research and knowledge management tool, NotebookLM already helped users organise ideas and connect documents. The new release shifts it from a reactive assistant into a more proactive research partner. Instead of requiring a fully prepared corpus, users can now start with rough questions and let NotebookLM propose sources, assemble a knowledge base, and iterate through analysis steps. For knowledge workers who routinely deal with long reports, messy datasets, and evolving questions, NotebookLM agentic research aims to shorten the path from raw material to structured insight.
Gemini 3.5 Upgrade and Advanced Reasoning for Researchers
At the core of the upgrade is the Gemini 3.5 upgrade, which powers a new chat experience focused on reasoning over complex, multi-step tasks. Gemini 3.5 runs together with Google’s Antigravity cloud infrastructure, yielding higher accuracy, reliability, and clearer explanations when summarising or comparing long documents. Internal evaluations show that the upgraded system achieved an average win rate of more than 65% across five core categories, with a 69.9% win rate in large-document analysis and 78.2% in advanced web research and source discovery. For researchers, this translates into better prioritisation of sources, more nuanced synthesis, and fewer hallucinated connections. The system can now explain how it arrived at an answer, giving analysts more confidence when using it for literature reviews, policy comparisons, or technical audits. In effect, the Gemini 3.5 upgrade turns NotebookLM into a reasoning engine built directly into a knowledge management tool.
Antigravity Cloud Code Execution and Software Skills
One of the most significant changes is Antigravity-based cloud code execution inside each notebook. Every project now includes a secure cloud computer that can write and run code for deeper quantitative analysis. NotebookLM ships with more than 100 curated software skills, so users can move from qualitative reading to programmatic analysis without leaving the workspace. Researchers might combine heterogeneous datasets, clean messy CSV files, run statistical tests, or prototype small simulations directly inside the notebook. Using cloud code execution, NotebookLM can automate repetitive data workflows, such as transforming formats, calculating derived metrics, or generating charts for inclusion in reports. This brings traditional coding tasks into the same environment as source reading, note-taking, and synthesis, turning NotebookLM into an end-to-end research environment instead of a passive document chat tool.
Agentic Workflows, Source Discovery, and Output Formats
Agentic research in NotebookLM is not only about better reasoning; it is also about how projects start and end. Previously, users needed to upload or paste their own sources before any meaningful analysis. Now, users can begin with loose ideas, and NotebookLM can help build a source repository directly in chat, find related materials, identify primary sources in multiple languages, and use Google Search for source discovery. Users stay in control of which sources enter the notebook, with clear attribution maintained throughout. On the output side, NotebookLM’s expanded formats cover PDF reports with charts and tables, DOCX, Markdown, TXT, CSV, JSON, XLSX, and visual assets like PowerPoint decks and images. Generated content can be edited through a studio panel, letting teams refine drafts into final deliverables. These capabilities turn agentic workflows into complete research pipelines.
Availability and Implications for Knowledge Workers
The upgraded NotebookLM agentic research experience is rolling out on the web to Google AI Ultra subscribers and Workspace business customers with AI Ultra Access and AI Expanded Access, with broader access planned. For now, this limits the upgrade to advanced users and organisations, but it also signals NotebookLM’s positioning as a professional-grade knowledge management tool. Knowledge workers who adopt the new system gain a unified space where AI research automation, source discovery, reasoning, and cloud code execution sit alongside familiar documents. They can convert complex specifications into readable guides, examine sales or operational data, and generate ready-to-share reports and presentations without juggling multiple apps. As availability expands, the combination of agentic workflows, Gemini 3.5 reasoning, and Antigravity cloud computing is likely to redefine how teams plan, execute, and share complex research projects.
