NotebookLM after the Gemini 3 update: what you can actually do with it

Put thirty PDFs, a few Google Docs and a YouTube video into one notebook, ask a question, and you get an answer with a pointer to the passage it came from. NotebookLM has done that since the start. In late 2025 so much was added that YouTube suddenly called the tool "insane": a new language model, tables you can export to Google Sheets, photos as sources and a research feature that finds sources on its own. In the video above, Julian Goldie's digital avatar walks through those features. Below is what actually changed, what it costs and what has happened since.

As of September 2026. This article first appeared in January 2026 and has been updated: since July 2026 NotebookLM is called Gemini Notebook.

What is NotebookLM?

NotebookLM is a Google research assistant that works with sources you provide: PDFs, Google Docs and Sheets, Word files, websites, YouTube videos, audio and images. Its answers are based on those sources and include citations to the passages they came from. From the same sources it also produces summaries, study material, audio and video.

The difference from a regular chatbot is that citation. For every answer you can check with one click whether it holds up. Do that: a citation makes checking easy, not unnecessary.

NotebookLM is now called Gemini Notebook

On 16 July 2026 Google renamed it Gemini Notebook. It is the same product: your notebooks stay, shared links redirect and it remains a standalone app. What's new is that you can also use your notebooks from the Gemini app, and that every notebook gets a secure cloud environment where it can write and run code for data analysis. That feature went first to Google AI Ultra and Workspace business customers, with Pro following later.

Google says more than 30 million people now use the tool. In the rest of this article I use the name NotebookLM, because the update shipped under that name.

Gemini 3: why the model underneath matters

In December 2025 NotebookLM moved from Gemini 2.5 Flash to Gemini 3. According to Google it was the most requested improvement, mentioned three times as often as anything else. Google promises better reasoning and better understanding of images and text together.

The difference should show most in questions that combine several sources ("where do these three reports contradict each other?") and in sources that aren't clean text, such as photos and charts. That second point became more relevant at the same time, because images were added as a source type in the same period.

Data Tables: from messy sources to a spreadsheet

NotebookLM pulls data out of your sources and puts it in a table, which you export to Google Sheets with one click.

Say you have forty completed feedback forms as PDFs. You ask for a table with, per form, the department, the main complaint and the suggested fix. Instead of an afternoon of retyping you get a spreadsheet you can sort and count. Do check a few rows against the originals: a table always looks more reliable than it is.

Data Tables arrived in December 2025, first for paying subscribers (Google AI Pro and Ultra). Free users now have access with limited usage.

Video Overviews and the Studio

Video Overviews turn your sources into an explainer video: slides with images, quotes and numbers from your material, narrated by an AI voice. In the video above, Julian Goldie shows how to use this for onboarding and team training.

The feature is older than the update suggested: Video Overviews launched in July 2025, together with the redesigned Studio panel. That is also where you create mind maps, reports and Audio Overviews, the podcast-style conversation about your sources. Audio Overviews now come in several formats, such as a brief summary, a critique and a debate. Later, "cinematic" Video Overviews followed, which use Gemini 3, Nano Banana Pro and Veo 3 to produce moving images instead of slides.

Photos of whiteboards and notes as sources

Since November 2025 you can add images as sources, and in December the mobile app gained a camera button. Photograph a whiteboard after a meeting or a page of handwritten notes, and it becomes a source you can ask questions about.

Flashcards and quizzes

Since September 2025 NotebookLM creates flashcards and quizzes from your sources. Upload a manual and you have practice questions for a new colleague. On every card or question you can click explain for a fuller answer with a citation to the source.

Deep Research: finding sources instead of only reading them

Until November 2025 NotebookLM only worked with what you put in. Since 13 November, Deep Research also searches the web itself. It drafts a research plan, browses hundreds of sites and delivers a report listing the sources it used. You add that report and its sources to your notebook with one click. There is a fast mode and a thorough one.

That shifts the tool's role from "read my documents" to "find the material, then read it". If source selection matters to you, look at what comes in. Deep Research picks the sites; you decide what stays.

What is free and what isn't?

The video calls NotebookLM "largely free". That holds for light use. These are the limits according to Google's help page:

The biggest differences are the number of sources per notebook (50 versus 300) and Deep Research (10 a month versus 20 a day).

Where to start

The video makes a fair point: most people try a tool like this once, get confused and give up. So don't start with an empty notebook; start the other way round. Pick one concrete question you have right now, such as "what do this quarter's customer calls say about pricing?". Add only the sources that deal with it, and only then ask for a table, a summary or a quiz.