Three years after launching as a Google Labs experiment, NotebookLM has quietly become one of the most practical AI research tools out there. Millions of people—students, researchers, professionals, and curious minds—use it daily to organize their thinking, spot connections across documents, and spark new ideas. But today, Google is giving NotebookLM its biggest upgrade yet, and it fundamentally changes what the tool can do.

NotebookLM now runs on Gemini 3.5 and Antigravity, Google’s agent-first AI-powered integrated development environment. This isn’t just a minor model swap. It’s a complete transformation that turns NotebookLM from a smart document reader into a full-blown research agent capable of tackling complex, multi-step projects.

The switch to Gemini 3.5 brings noticeably more accurate and reliable information, plus better visibility into how the AI thinks through problems. In Google’s own side-by-side evaluations against the previous system, the upgraded NotebookLM won over 65% of the time across five core evaluation dimensions—that’s a 15-point margin above parity.

The improvements are even more striking in specific areas. For large document analysis, NotebookLM achieved a 69.9% win rate. For advanced web research and source discovery, it hit 78.2%. Those aren’t incremental gains. They’re the kind of jumps that make you trust the tool for actual work instead of just casual exploration.

But the real magic comes from Antigravity. Every notebook now comes equipped with its own secure cloud computer. This means NotebookLM can actually write and run code on your sources instead of just reading them back to you. It’s not summarizing anymore—it’s analyzing, computing, and building things directly from your research material.

The system includes more than 100 curated software skills that unlock capabilities ranging from data analysis to code execution to generating different output formats. You can create workflows inside your notebooks that previously required switching between multiple applications.

You can now start research from a blank notebook

Before this update, NotebookLM was most helpful when you already had sources and a clear project direction. You’d upload documents, ask questions, and get answers. That’s still true, but now you can start with just a loose idea or a question.

NotebookLM can guide you through building your source repository directly in chat. It uses its own research skills and Google Search to suggest useful sources, find primary materials in other languages, discover related works by authors you’re exploring, and add selected materials straight into your notebook.

You’re still in control of what gets added. Every source is clearly attributed, so your work stays grounded in information you trust. This makes it way easier to get research started without the awkward “what do I even upload first?” moment.

More output formats, more control

The upgrade also expands what NotebookLM can create and export. You can now ask it to produce outputs in many more formats and give detailed instructions to shape how they come out. Plus, you can edit outputs after they’re generated.

New output formats include:

• Data visualizations and charts (PNG, SVG)
• Documents (PDFs, DOCX, Markdown, text files)
• Images with Nano Banana (PNG, JPG, GIF)
• Structured data (CSV, JSON)
• Microsoft Excel (XLSX)
• Microsoft PowerPoint (PPTX)

So if you’re a data analyst combining conflicting datasets from different countries, you can ask NotebookLM to find web context, write code to analyze the data accurately, and create charts plus a PDF report showcasing everything. A program manager deciphering complex customer integration specs can transform technical documentation into a simplified guide, slide deck, and step-by-step roadmap. A gym owner running a media campaign can analyze raw sales data against ad spend and calculate the campaign’s financial impact to decide whether to expand to other cities.

Another important change is transparency. NotebookLM now shows detailed steps in the chat about how it arrived at answers, so you can check the output and understand the reasoning process. This “visibility into the thinking process” is huge for building trust, especially when you’re using the tool for actual work rather than just casual browsing.

Who can use this right now

The updates are rolling out globally on the web starting today, but there’s a catch: they’re currently available only to Google AI Ultra subscribers and Workspace business customers with AI Ultra Access or AI Expanded Access.

Google says it plans to expand access to others over time, but if you want to try this right now, you’ll need one of those subscriptions. That’s pretty typical for major Google AI releases—they often start with Ultra or enterprise users before going broader.

NotebookLM’s Gemini 3.5 and Antigravity upgrade is one of the most significant improvements to Google’s research tool since its launch. It’s more accurate, more transparent, more capable, and way more practical for real work. The secure cloud computer, 100+ software skills, expanded output formats, and chat-based source discovery transform it from a document reader into a comprehensive research agent.

If you’re a Google AI Ultra subscriber or have Workspace AI Ultra Access, the upgrade is available now. And if Google follows its typical pattern, broader access will likely come eventually.

For researchers, content creators, students, and anyone who does serious work with information, this upgrade makes NotebookLM worth revisiting—even if you tried it before and felt it was limited. The tool has fundamentally changed, and it’s changed in the right direction.