What the Gemini 3.5 Upgrade Means for NotebookLM
NotebookLM is Google’s source-grounded document analysis AI that lets users upload and question PDFs, web pages, media, and office files, then turn those materials into organised insights, structured notes, and exportable outputs that support research, study, and professional knowledge work. With the new NotebookLM Gemini 3.5 upgrade, the app shifts from a smart summariser into a more active research environment. The chat experience now runs on Gemini 3.5 together with Google’s Antigravity tool layer, giving the system expanded reasoning steps that are shown directly in the conversation. Google frames this as a response to demand for clearer, more transparent thinking traces in AI research tools. Company evaluations claim an average win rate above 65 percent over the previous NotebookLM model, with especially strong gains in large document analysis and advanced web research.
Agentic Research Features: From Loose Ideas to Curated Sources
The standout change is a new set of agentic research features that turn NotebookLM into a more proactive research assistant. Instead of starting only from uploaded files, users can begin with a question, then let NotebookLM tap Google Search to suggest relevant web sources. Researchers can choose which of these suggested pages, papers, or videos become part of a notebook, keeping work grounded in selected material while reducing set-up effort. This source discovery layer can surface primary sources, related work by the same author, or materials in other languages that might be missed in a manual search. According to WinBuzzer, NotebookLM now supports “advanced web research and source discovery” that scored a 78.2 percent win rate over the prior system, indicating stronger performance when building and refining a research corpus around open-ended topics.
Cloud-Based Code Execution and Curated Software Skills
NotebookLM now embeds a secure cloud computer in every notebook, turning it into something close to a code execution notebook for research workflows. This environment can write and run code directly from the chat interface, allowing users to analyse datasets, test algorithms, or automate repetitive calculations without leaving the app. Google says the system includes more than 100 curated software skills, which help the document analysis AI interact with the sources inside a notebook and run the right tools for tasks such as statistical checks, text parsing, or basic simulations. For students learning to code, this means they can move from theory in a PDF to live experiments in the same workspace. For professionals, it reduces the need to jump between AI chat, an IDE, and a separate analytics tool to validate findings or prototype ideas.
Downloadable Reports, Charts and Multi-Format Outputs
On the output side, NotebookLM’s Gemini 3.5 integration turns chat results into a wider range of finished artefacts suitable for classes, labs, or meetings. Users can instruct the system to generate reports, charts, slide decks, spreadsheets, images, structured datasets, and more, then download them from a studio-style panel. Supported formats include PDF, DOCX, PPTX, XLSX, CSV, JSON, markdown, PNG, SVG, JPG, GIF, and plain text files, meaning the same research session can yield a polished presentation plus underlying data files for further work. NotebookLM also keeps clear attribution to the sources inside each notebook, so students and researchers can trace claims back to their origin documents. Combined with visible reasoning steps, this makes the tool better suited to academic and professional environments that require explanation, not only answers.
Who Gets Access First and How Workflows Change
For now, the most advanced NotebookLM Gemini 3.5 and Antigravity capabilities sit behind Google’s premium AI Ultra subscription and selected Workspace plans with AI Expanded access. Eligible school teams, universities, research groups, and workplace users will be the first to test the upgraded document analysis AI in their day-to-day workflows, with Google stating that broader availability is planned over time. In practice, the new agentic research features let students and professionals move from a loose question to curated sources, deeper calculations, and exportable outputs without hopping between tools. A single thread might start with a topic prompt, expand into source discovery and reading, run code to check a result, and end with a downloadable chart or slide deck. For knowledge workers managing complex documents, this positions NotebookLM less as a passive summariser and more as an integrated research partner.
