What NotebookLM Is Now: From AI Note-Taking to Research Partner
NotebookLM is an AI note-taking and research tool that now combines Gemini 3.5, code execution, and agentic research workflows so knowledge workers can move from raw sources to fully formatted outputs without leaving a single workspace. Built originally as an experimental product to help people understand their documents, it has grown into a collaborative research hub used by individuals and teams to organize ideas, surface connections, and spark new insights. Users can upload PDFs, websites, YouTube videos, audio files, Google Docs, and Slides to form a shared knowledge base. The latest upgrade changes not only how information is summarized, but how complex projects are conducted: instead of manually juggling separate tools for analysis, coding, and reporting, NotebookLM now centralizes these capabilities, turning chat-based prompts into structured, repeatable research workflows that can scale from a single paper to large document collections.
Gemini 3.5 and Antigravity: A New Core for AI Research Tools
NotebookLM now runs its chat experience on Gemini 3.5 and Antigravity, aligning it with Google’s latest model stack for AI research tools. This matters because reasoning quality and transparency sit at the center of serious research workflows. The system now shows expanded thinking steps directly in chat, giving users insight into how conclusions were reached instead of hiding the chain of reasoning behind a single answer. According to Google, “the upgraded NotebookLM achieved an average win rate of over 65% across our top five core evaluation dimensions,” with a 69.9% win rate in large document analysis and 78.2% in advanced web research and source discovery compared with the prior baseline. For knowledge workers, this translates into more reliable answers on long reports, policy documents, literature reviews, and technical specs, with clearer links back to the supporting sources inside each notebook.
Code Execution Notes: Secure Cloud Computers Inside Each Notebook
The most transformative change is the addition of a secure cloud computer attached to every notebook, turning NotebookLM into more than a summarizer. Within the same interface where you chat and take notes, NotebookLM can now write and run code against your sources, enabling code execution notes for data-heavy work. This secure environment taps into more than 100 curated software skills, including tools for data analysis, visualizations, and other advanced workflows. Instead of exporting tables to another app, running scripts, and re-importing results, users can perform deeper research and complex analysis in place. For example, a researcher can run statistical code over a dataset, generate charts from survey results, or prototype simulations directly from notebook context. Because the compute environment is bound to each notebook, source control and reproducibility become easier: the code, outputs, and references all live alongside the underlying documents.
Agentic Research: From Questions to Source Libraries and Structured Reports
NotebookLM’s new agentic research tools change how projects start and evolve. You no longer need a pre-built pile of documents; you can begin with an idea or a question, then let NotebookLM help assemble a source library while keeping you in control of what gets added and cited. In chat, the system can search for relevant materials, show reasoning steps, and perform multi-step tasks such as comparing sources, outlining arguments, or synthesizing themes. It becomes an active collaborator that “builds a source repository in chat” rather than a passive notepad. These agentic capabilities suit knowledge workers running literature reviews, competitive analyses, or multi-stakeholder reports, where the workflow includes discovery, triage, summarization, and synthesis. The emphasis on attribution means that even as it automates these steps, users can still trace each claim back to its supporting document or web source within their notebook.
From Outputs to Outcomes: Multi-Format Reports and Research Workflows
NotebookLM now closes the loop from research to deliverables by generating downloadable artifacts in the formats knowledge workers use every day. From the studio panel, users can export reports, charts, slide decks, spreadsheets, structured data files, and images, shaping outputs with detailed instructions and editing them afterward. Supported formats span PDFs, DOCX, XLSX, PPTX, CSV, JSON, markdown, text, PNG, SVG, JPG, and GIF, with the ability to turn a notebook’s contents into full PDF reports complete with charts and tables. This is where AI note-taking evolves into a full research hub: a single notebook can hold sources, AI-assisted analysis, code execution notes, and final outputs ready for stakeholders. Multilingual workflows add another layer, allowing instructions in one language and outputs in another, which is particularly helpful for cross-border collaborations or research projects involving non-native-language sources.
