What NotebookLM’s Gemini 3.5 Upgrade Changes
NotebookLM is Google’s AI note-taking assistant that helps users collect sources, understand long materials, and turn fragmented notes into structured analysis, and its Gemini 3.5 upgrade aims to make that entire research flow faster, clearer, and easier to share. The service now runs on the NotebookLM Gemini 3.5 model combined with the Antigravity platform, which Google says brings “even more accurate and reliable information along with better visibility into the thinking process.” In side-by-side internal tests, the refreshed system achieved an average win rate of more than 65% across five core evaluation categories compared with the previous version. Performance in large document analysis reached 69.9%, while advanced web research and source discovery hit 78.2%, showing clear gains for anyone working with complex sources. For now, the upgraded AI research tools are rolling out to Google AI Ultra subscribers and Workspace business users with AI Ultra or AI Expanded Access.
Cloud Code Execution Turns Notes Into Live Analysis
The headline functional change behind the NotebookLM Gemini 3.5 shift is cloud code execution. Each notebook now includes a secure cloud computer that can run code directly inside the workspace. That means users can move from reading a dataset in their notes to testing a Python snippet or exploring a small model without leaving their note-taking assistant. According to NewsBricks, the environment ships with more than a hundred curated software skills, so the system can call specialised tools for statistics, text processing, or data cleaning as part of multi-step research flows. For data-heavy projects, this blurs the line between notebook and lightweight lab environment: you can draft ideas in natural language, have the model propose code, execute it in the cloud, then feed results back into your analysis, all in one continuous thread.
PDF Export and Rich Formats Make Sharing Research Easier
NotebookLM’s new PDF export feature pushes it beyond a chat-style note-taking assistant into a publishing-ready workspace. From the Studio Panel, users can now download outputs in widely supported formats including PDF, DOCX, Markdown, TXT, PNG, SVG, JPG, GIF, CSV, JSON, XLSX, and PPTX. Google highlights that the upgraded format supports “PDF reports with charts and tables,” so structured outputs like research summaries, slide decks, or data visualisations can move straight from NotebookLM into team workflows or archives. Crucially, outputs are editable after generation: you can request changes, refine sections, and only then export the final version. For teams, this means the same space that hosts sources and analysis can also produce finished documents without manual copy-paste into other tools, turning NotebookLM into both a drafting room and a packaging layer for research.
Smarter Source Discovery and Research Workflows
Alongside the NotebookLM Gemini 3.5 engine, Google has focused on how sources enter the notebook in the first place. Users can still upload files or use search, but they can now drop an idea or question into the chat between the Sources and Studio panels and let NotebookLM propose a reading list. The assistant can tap Google Search to find relevant, high-quality sources and help uncover primary materials in other languages or related works by the same author, then let you add or discard items one by one. On top of that, the new agentic capabilities are designed for complex, multi-step research flows: summarising large documents, comparing viewpoints, extracting quantitative data, and then feeding those findings into structured outlines or reports. The result is a tighter loop from question, to source discovery, to analysis, to shareable output.
From Note-Taking Assistant to Complete Research Platform
Taken together, the Gemini 3.5 upgrade, cloud code execution, and rich export options reposition NotebookLM from a smart notepad to a more complete research and productivity platform. It now helps at every stage: building a source repository, understanding long materials, running computations, and exporting polished outputs to PDF, slides, or data formats. For students and researchers, this can cut down on context-switching between chatbots, coding environments, and office suites. For teams, it offers a shared space where background research, exploratory analysis, and final documents live side by side. Availability is currently limited to AI Ultra and certain Workspace tiers, but the direction is clear: NotebookLM is evolving into an AI research hub where reasoning, code, and collaboration are tightly linked instead of scattered across separate tools.
