01What Is NotebookLM Research Agent?
NotebookLM research agent is a new feature in Google's NotebookLM that lets you ask a loose question and it finds relevant sources from the web for you. It can read the web, pull in high-quality sources, and add them to your notebook. This makes research faster and more reliable.
In this episode, Julian Goldie explains that the update gives NotebookLM full agentic capabilities. That means it can research, code, run code, generate videos, and more. It works like a research assistant that handles the hard work of finding and organizing information.
The agent also has a deep research mode. You can switch between fast research and deep research. Deep research takes longer but gives more in-depth results.
The new system has a 78.2% win rate against the old one, so it is a big step up.
02Main Features of the NotebookLM Update
The update adds several powerful features. Here are the main ones:
- Agentic research mode – Start with a question and NotebookLM finds sources for you using Google search.
- Advanced reasoning – More powerful reasoning in chat, so you can have deeper conversations with your notebook.
- Code execution – You can write and run code directly in NotebookLM, create spreadsheets, search across documents, and make charts.
- New output formats – Generate videos, slide decks, quizzes, mind maps, infographics, flashcards, and full research reports.
- PDF reports – The research agent can compile findings into a beautiful PDF with charts and data.
These features make NotebookLM a one-stop tool for turning ideas into finished products. Julian shows examples of research reports, infographics, and videos all made in NotebookLM.
03How to Get Started with NotebookLM Research Agent
Getting started with the new NotebookLM research agent is simple. Follow these steps:
- Open NotebookLM and create a new notebook or select an existing one.
- Click on the research option in the chat.
- Choose between fast research and deep research mode.
- Type your question or loose idea in the chat.
- Let the agent search the web and pull in relevant sources.
- Review the sources and add them to your notebook.
- Chat with your notebook to ask follow-up questions.
- Use the Studio to generate videos, podcasts, or other output formats.
The process is designed to be quick. You can go from a question to a researched visual in minutes. Julian notes that you don't need to have sources set up beforehand; the agent finds them for you.
04How to Use NotebookLM Research Agent for Deep Research
In the episode, Julian shows how to use the research agent to create a deep research report. First, you start with a question. For example, you might ask about AI agents in your industry.
The agent then goes out and finds the best sources, not just the first 10 Google links.
It pulls those sources into your notebook and summarizes them. You can then chat with your notebook to ask for more details or clarification. Every answer cites its source, so you know where the information comes from.
After the research is done, you can visualize it. You can turn the research into a mind map, infographic, slide deck, or even a podcast. Julian shows an example of a full research report with charts and data.
He also mentions that you can create PDF reports to share.
For coding, you can ask the agent to create a spreadsheet or run code. It can also search across your documents to find specific information.
05Results and Comparisons from the Episode
The episode compares the new way of researching with the old way. The old way involved opening 20 tabs, using outdated sources, copying and pasting, and losing links. The new way is one question, one button, and everything is organized.
Julian highlights that the new system has a 78.2% win rate against the old version. That means it performs better in most tests. He also shows that the research agent can generate a beautiful research report in one click.
You can watch the agent find sources in real time.
Another result is that all your research is stored in one place. You can come back to it later. This makes it easier to use in the future, unlike the old way where you might forget where you put things.
Julian also demonstrates connecting NotebookLM to an agent operating system. This lets you control it alongside other agents like Hermes and OpenCore. He says this makes NotebookLM 10 times more powerful because you can automate tasks and distribute content easily.
06Tips, Limits, and Mistakes to Avoid
Julian offers several tips for getting the most out of NotebookLM. First, use the deep research mode when you need thorough results. Fast research is good for quick questions, but deep research gives more depth.
Second, don't keep NotebookLM in isolation. Connect it to an agent operating system using MCP (Model Context Protocol). This lets you use the research in other tools like SEO tools or video editing agents.
Julian says most people only create something in NotebookLM and don't plug it into anything else. But you should make your research actionable and distribute it.
One limit is that the update is rolling out first to Google AI Ultra members. Free users may have to wait. However, Julian expects it to come to everyone later, based on past patterns.
Another mistake is not storing your research in a knowledge base. Julian recommends using Obsidian to visualize your research as a knowledge graph. This helps you see connections and manage information better.
07Who Should Use NotebookLM Research Agent and Why It Matters
NotebookLM research agent is for researchers, technical professionals, and small business owners. Julian gives an example of a gym owner who could use it to analyze campaign data and make better decisions. You don't need to be a tech expert.
It matters because AI can now research for you. You just ask one question, and you get cited sources, a chat interface, and visual outputs. This saves hours of manual research.
You can turn a single idea into a report, podcast, slide deck, or video. Everything is stored in one place, so you never lose track of your work. This is especially useful for content creators, marketers, and anyone who needs to produce research-based content regularly.
Julian emphasizes that the new system is not just a one-off tool. It becomes part of a larger agent ecosystem. You can have multiple agents working together, like having a team of AI assistants.
This makes research and content production much faster and more efficient.
