What if you could begin a research project with just one question and end up with a report, charts, and a presentation ready to share? Google is expanding NotebookLM to turn it into a more autonomous research companion, capable of analyzing documents, searching for sources, and generating different types of files.
The update, presented by Google, adds new reasoning features and tools for more complex research projects. The improvements have already started rolling out on the web to Google AI Ultra users and Workspace business customers with AI Expanded Access.
NotebookLM offers a more powerful chat
NotebookLM now runs on Gemini 3.5 and Antigravity, according to Google. The company says this combination improves response accuracy, information reliability, and visibility into how the system arrives at its results.
One of the most important new features is that each notebook has a secure computer in the cloud. This allows NotebookLM to write and run code for deeper analysis, work with large volumes of information, and handle tasks that previously required several separate tools.
The system also includes more than 100 selected software skills to expand its capabilities. In practice, this can help compare data, detect patterns, process lengthy documents, or prepare results with a clearer presentation.
Google says that, in benchmark evaluations against the previous version, the new NotebookLM achieved an average preference rate of more than 65% across five main dimensions. It reached 69.9% in large-document analysis, while advanced web research and source discovery reached 78.2%.
These figures come from Google's internal evaluations against its previous system, not from an independent comparison with all the research tools available.
More formats for turning research into results
NotebookLM is no longer limited to answering questions about uploaded sources. It can now turn that information into finished products and let users provide more detailed instructions about how they want to receive them.
Available formats include:
- Charts and data visualizations in PNG and SVG.
- Documents in PDF, DOCX, Markdown, and text.
- Images created with Nano Banana in PNG, JPG, and GIF.
- Structured data in CSV and JSON.
- Microsoft Excel spreadsheets in XLSX.
- Microsoft PowerPoint presentations in PPTX.
For example, you could upload sales data, request an analysis of monthly trends, and receive a PDF report with tables and charts. You could also turn technical documentation into a simple guide, a presentation for the team, and a step-by-step roadmap.
Google also says that generated results can be edited after they are created. This matters because AI can prepare a useful first draft, but human review is still necessary to correct details, adjust the tone, and validate the conclusions.
Research can start with an idea
Until now, NotebookLM was especially useful when users already had documents and knew how to organize their project. With this update, the tool can also help build a source base from the beginning.
You can start with a broad question or a loosely defined idea. NotebookLM can guide the search for materials, locate primary sources in other languages, and find papers related to an author or topic. It can also use Google Search to discover relevant information on the web and add it to the notebook.
Users retain control over the sources they add. Google says that results remain clearly linked to their references, a key feature for checking where each claim comes from.
Does this mean NotebookLM replaces the researcher's judgment? No. Its value lies in speeding up exploration, organizing scattered information, and preparing drafts that can be analyzed. Source selection, fact-checking, and final decisions still depend on the person.
Use cases for work, study, and business
The new features can be useful in many different situations:
- Research and analysis: an analyst can combine data from several countries, correct formatting differences, search the web for additional context, run code to analyze the information, and create a report with visualizations.
- Technical teams: a project manager can turn complex integration specifications into a simplified guide, a presentation, and a work plan for the team.
- Small businesses: a gym owner can compare sales with advertising investment, calculate a campaign's financial impact, and decide whether expanding to other cities makes sense.
- Education: a teacher or student can turn study materials into quizzes, worksheets, summaries, and structured documents.
The central idea is simple: NotebookLM is trying to reduce the distance between having information and being able to use it. Instead of jumping between a search engine, a spreadsheet, a document editor, and a presentation tool, users can coordinate several stages from one workspace.
These capabilities are rolling out globally on the web for Google AI Ultra users and Workspace business customers with AI Expanded Access. Google plans to extend them to more users over time.
The update shows where AI applied to work is heading: fewer isolated answers and more complete processes. Even so, the quality of the result will depend on the sources you choose, the instructions you provide, and the review you conduct before making a decision. AI can help you research better, but you are still the one who must decide what deserves your trust.