Google announced an update to its AI research tool, NotebookLM, switching the default model to Gemini 3.5 and adding a new ability to gather information from chat content. This changes how the product is used, freeing users from relying entirely on pre-uploaded materials to build their knowledge bases.
Default model upgrade
With this update, NotebookLM uses Gemini 3.5 as the default model. Google has also added new software capabilities to help users generate different types of output during the research process, further enhancing its organization and production capabilities beyond question answering.
Support building databases from chat
NotebookLM previously required users to prepare materials first, then extract information and draw conclusions based on them. Now, users can give more detailed instructions directly in the dialog box, allowing the system to generate content, and continue to modify it after generation, gradually accumulating it into a usable database.
Export formats have been further expanded.
The new version supports exporting various file types, including charts and data visualizations, PDFs, .docx files, Markdown files, text files, and structured data formats such as .csv and .json.
- Chart formats include .png and .svg.
- Document formats include PDF and Word files.
- It also supports exporting to Excel and PowerPoint.
Google also mentioned that NotebookLM can generate image content powered by Nano-banana, making it easy for users to directly convert research results into presentation materials or visualizations.
Show the generation process in the chat
Google introduced the "Deep Research" mode for NotebookLM last year for more structured online research. With this update, the system will display more detailed steps in the response generation process within the chat interface, making it easier for users to verify the basis of their output.
Google says these features are available to Google AI Ultra users and Workspace enterprise customers who have activated AI Ultra Access and AI Expanded Access, starting today, and will be expanded to more users in the future.
