What NotebookLM’s Gemini 3.5 Upgrade Actually Is
NotebookLM’s Gemini 3.5 upgrade is a major update to Google’s AI research tool that turns static notebooks into active research agents able to find sources, run code, and generate downloadable reports from a secure cloud computer. Built on Gemini 3.5 and Google’s Antigravity layer, the new NotebookLM chat experience aims to give more accurate, consistent answers while showing the reasoning steps that led there, so users can inspect how conclusions were reached. Google describes NotebookLM as a “collaborative research and knowledge platform” that millions already use to organize information and uncover connections across documents. With this release, the product expands from summarizing and comparing uploaded sources to initiating research from loose questions, pulling in web materials, and transforming them into structured outputs for teams, students, and researchers who need deeper analysis rather than isolated AI replies.
Advanced Reasoning and Agentic Capabilities Change the Workflow
The headline change in NotebookLM Gemini 3.5 is agentic capabilities: the system can now plan multi-step research tasks and call tools on its own. Instead of requiring a clean set of pre-uploaded PDFs before anything useful happens, users can start with a vague topic and use Google Search inside NotebookLM to discover relevant web sources, then selectively add them into a notebook. Source suggestions can surface related work, primary materials in other languages, and writings by the same author that might otherwise be missed. Google says NotebookLM now includes Deep Research stages and visible answer steps that aim to provide “even more accurate and reliable information along with better visibility into the thinking process.” In practice, this shifts NotebookLM from a glorified summarizer into an AI research tool that helps build, curate, and interrogate a live source library.
Code Execution AI: From Notes to Computation in One Place
Every notebook now includes a secure cloud-based computer, which is where NotebookLM’s code execution AI features appear. Antigravity 2.0 provides an agentic coding layer that can write and run code to analyze data, test ideas, or transform raw research material into structured outputs. Google says NotebookLM can tap into more than 100 curated software skills, covering tasks like statistical analysis, data cleaning, and visualization. In practical terms, a single question can lead to automatic source gathering, a custom calculation, and a formatted answer without jumping between a browser, a coding environment, and a document editor. Internal evaluations reported an average win rate above 65 percent versus the previous version, including “a 69.9% win rate in large-document analysis and a 78.2% win rate in advanced web research and source discovery,” suggesting the new workflow is not only more automated but measurably stronger at handling complex material.
Downloadable Reports and Rich Exports Make Results Reusable
NotebookLM’s expanded export options make its research automation useful beyond the chat window. The system can turn a notebook into PDF reports with charts and tables, detailed budget-style summaries, or structured datasets for further analysis. Supported formats now include documents (PDF, DOCX, Markdown, TXT), structured data (CSV, JSON), spreadsheets (XLSX), slide decks (PPTX), and visual outputs such as charts and images in PNG, JPG, GIF, and SVG. For more visual tasks, NotebookLM can even generate images using Nano Banana. Crucially, exports are editable: users can give instructions before creation, then ask the AI to revise a report, adjust chart selections, or reframe a slide deck after seeing the first version. This makes NotebookLM feel less like a chatbot and more like a flexible report generator that can continuously refine outputs until they match a specific research brief.
Who Gets Agentic NotebookLM First—and Why It Matters
The most advanced agentic capabilities in NotebookLM Gemini 3.5 will not reach everyone at once. Google AI Ultra subscribers and eligible Workspace customers are the first to gain access, with broader availability promised later. For these early users, NotebookLM now spans the full research pipeline: starting from loose questions, discovering and curating sources with Google Search, running code in a secure cloud computer, and exporting polished artifacts in many formats. That end-to-end flow turns NotebookLM from a note-taking companion into an active research agent that can reduce context switching between tools and cut down manual data wrangling. For teams already living in Google’s productivity stack, the upgrade positions NotebookLM as a central AI research tool—a place where background reading, exploratory analysis, and final deliverables all share the same source-grounded context instead of being scattered across separate apps.
