What the NotebookLM Gemini Upgrade Actually Is
The NotebookLM Gemini upgrade is a major update to Google’s source-grounded research notebook that combines the Gemini 3.5 model, Antigravity reasoning tools, and cloud-based code execution to turn document analysis, web research, and results export into a single integrated workflow. Originally launched as an experimental Labs project for understanding topics, NotebookLM has grown into an AI research tool that helps organise ideas, connect documents, and generate insights. With Gemini 3.5 now powering responses, the system focuses on harder, multi-step research flows rather than casual chat. Google says performance against the prior NotebookLM has improved with an average win rate above 65 percent across five evaluation categories, including 69.9 percent for large document analysis and 78.2 percent for advanced web research and source discovery, although these results come from internal tests rather than independent benchmarks.
Antigravity Reasoning and Cloud Code Execution
Antigravity is the new execution layer that turns NotebookLM into more than a reader of sources. Each notebook now includes a secure cloud computer that can write and run code from inside the workspace, so a single research question can move from source selection to calculation to a finished output without switching tools. Antigravity 2.0 adds agent-like behaviour: NotebookLM can call curated software skills, inspect material, run checks, and produce structured artifacts backed by the original sources. This cloud code execution is central for data-heavy work, from basic statistics to more complex analysis pipelines, while keeping the notebook as the control panel. For research teams comparing AI research tools, this combination of Antigravity reasoning and remote compute makes NotebookLM feel closer to a lightweight, guided lab environment than a standard chatbot or standalone coding editor.
New Source Discovery, Deep Research, and Visibility
The upgrade also changes how work starts. NotebookLM no longer depends on fully prepared inputs: users can begin from loose questions, then use integrated Google Search to discover sources and choose what enters a notebook. Source suggestions can surface primary material in other languages or related work from the same author, widening coverage for literature reviews and background scans. Deep Research plays a role in the flow, helping with multi-hop reasoning over web and file sources without losing traceability. NotebookLM adds clearer answer-step visibility, giving users a better view of how the AI reached a conclusion. More than 100 curated software skills sit behind this deeper analysis, helping with tasks like data extraction, transformation, and validation. Together, these updates push NotebookLM further toward end-to-end research workflows while staying grounded in explicit, user-selected sources.
Expanded Exports and Research Deliverables
NotebookLM’s Gemini 3.5 integration also expands what researchers can export. Outputs now include PDF reports with charts and tables, Word documents, Markdown and plain text files, plus CSV, JSON, and Excel spreadsheets. According to Google’s product descriptions, NotebookLM can also create PowerPoint-style slide decks, images, and structured data files from the same notebook session, turning source-backed analysis into ready-to-share deliverables. A studio panel lets users refine instructions and regenerate sections without starting from scratch, so long reports or data tables can be iterated quickly. Because outputs are grounded in user sources, they fit audit-heavy environments such as academic work or internal knowledge bases. This export flexibility, combined with cloud code execution, means NotebookLM can operate as a light publication pipeline: run the analysis, keep the traceable steps in one place, then generate the format the team or supervisor needs.
Who Gets Access First—and How It Shifts Research Workflows
The Gemini 3.5 NotebookLM upgrade is not available to everyone yet. Early access goes to Google’s premium AI Ultra subscribers and eligible Workspace business customers, with a wider rollout promised later. For those users, the change is less about raw model power and more about workflow consolidation: source discovery, Antigravity reasoning, cloud code execution, and exports now live in one notebook instead of several disconnected tools. This positions NotebookLM as a more powerful alternative to traditional research workflows that jump between search engines, coding notebooks, and office suites. Researchers already comparing AI research tools such as Claude, Perplexity Spaces, Elicit, AnythingLLM, or Copilot Notebooks will see NotebookLM move closer to an all-in-one environment. The main trade-off is access limits today, but if Google widens availability, this upgrade could reset expectations for what a research notebook should include by default.
