How to Use NotebookLM (Gemini Notebook): 2026 Guide

Last checked: 2026-11-04. The rename from NotebookLM to Gemini Notebook happened in July 2026; feature details below follow Google’s announcements through October 2026.

Learning how to use NotebookLM starts with a simple idea: instead of an AI that answers from its training data, you hand it your own documents and it answers from those, with citations pointing back to the source. Google renamed NotebookLM to Gemini Notebook on July 16, 2026 — the same standalone product, now tied more tightly into the Gemini ecosystem. The Standard plan is free. You add sources — PDFs, Google Docs, websites, YouTube videos, audio files — ask questions and get answers built only from what you uploaded. This gemini notebook guide covers setup, questions, Studio outputs, the October 2026 updates, and free-plan limits.

What NotebookLM is — and why it is now called Gemini Notebook

NotebookLM began life as Project Tailwind, an experiment Google showed at I/O 2023, and grew into a source-grounded AI research assistant: you upload your own material — PDFs, Google Docs, websites, YouTube videos, audio files — and it answers your questions using only those sources, linking every claim back to the passage it came from. On July 16, 2026, Google renamed it Gemini Notebook, describing it as a standalone product that remains “your premier research tool”, with the new name signalling deeper integration across the Google ecosystem (announced on the Workspace Updates blog the same day). Your old notebooks stayed put, and old notebooklm.google links redirect to the new canonical address, notebook.google.com. Google said at the time that more than 30 million people and 600,000 organisations were using it.

The rename brought two practical changes. Notebooks now appear inside the Gemini app with two-way sync, so work carries over between the two. Each notebook also gets a secure, isolated execution environment that can write and run code for data analysis on your sources — handy when your documents include spreadsheets or datasets. For a parallel story on the model side, there’s the Gemini 4 Carbon write-up.

How to use NotebookLM: set up your first notebook in five minutes

Getting from zero to a working notebook takes about five minutes:

1. Sign in with a Google account at notebook.google.com — a free account is enough.
2. Create a notebook. Click “New notebook” to open a blank one.
3. Add sources. Use the Sources panel to upload PDFs, connect Google Docs, paste website links, add YouTube videos, or upload audio files.
4. Let them process. NotebookLM indexes each source; large documents take a little longer.
5. Ask your first question. Type in the chat panel — and start with something whose answer you already know, so you can check the citations land where they should.

Know the free limits before you bulk up: 100 notebooks per account, up to 50 sources each, and each source can hold up to 500,000 words or 200MB.

Ask questions and get answers with citations

The chat panel is where most of the work happens. Ask in plain language — “What does the report say about Q3 revenue?” — and the answer is built from your uploaded sources, not the open web. Every factual claim carries a citation; click it and you land on the exact passage, highlighted. Follow-up questions keep the context, so you can drill down: “And how did that compare with Q2?” Because the tool is fenced inside your documents, specific questions against solid sources produce the best results. Vague questions over thin sources give vague answers — the quality of the output tracks the quality of the input.

Turn your sources into audio, video, slides, and mind maps

The Studio panel turns a notebook into shareable formats (the full feature list lives in the Gemini Notebook Help hub). Audio Overviews generate a podcast-style two-host discussion of your material. Video Overviews come in Explainer, Cinematic, and ~60-second Short formats — the Short format was announced by Google’s education team on October 16, 2026. Slide Decks export to PPTX and PDF and have been around since February 2026. You can also generate Mind Maps, quizzes, flashcards, and reports, and run Deep Research from the Sources panel; interactive overviews and real-time conversations arrived in mid-September 2026.

On October 19, 2026, Google added a co-creation workflow to the Studio: NotebookLM suggests topics drawn from your sources, you steer the content, format, and style, and a new Review toggle lets you check outline notes (for slides) or producer notes (for video) before anything renders. Google said the rollout would reach everyone globally within a week. Around the same stretch, the mobile app gained live audio recording of up to an hour (October 27, 2026) and a Connected App link inside the Gemini app (October 23, 2026).

Free-plan limits in 2026

Capacity on the free Standard plan is straightforward: 100 notebooks, 50 sources per notebook, 500,000 words or 200MB per source.

Plan	Notebooks	Sources per notebook
Free	100	50
Plus	200	100
Pro	500	300

Usage limits changed on September 2, 2026: Google replaced fixed daily counts with compute-based limits, with the quota refreshing periodically up to a weekly cap. Google doesn’t publish the exact numbers and they change — check the plans page for the current state. One detail carried by the limits guides: free-tier Video Overviews carry a visible watermark, removed on Pro and Ultra.

How it differs from ChatGPT

ChatGPT answers from its training data. NotebookLM answers only from the documents you upload, and every answer links back to the source passage it came from. That single difference shapes what each tool is good for — three short sketches make it concrete.

Exam prep. A student uploads lecture slides and two textbook chapters, generates a quiz from them, and plays the Audio Overview on the bus. Every practice answer cites the slide it came from, so revision time goes to the material, not to hunting through files.

Research and reporting. A journalist drops interview transcripts and background papers into one notebook, then asks questions that cut across them. The citations tell her exactly which transcript supports each quote — that is the difference between a usable quote and an unverifiable one.

Book notes. A reader uploads a novel and his own reading notes, generates a Mind Map, and asks how a theme in chapter 3 connects to chapter 7. The answer is grounded in the text he read, not a generic summary from a model’s memory.

Limitations to know before you rely on it

Answers are only as good as the sources you add — a thin or biased source set produces thin or biased answers, however polished they sound. Check the citations on anything you quote or publish. Feature rollout timing varies by account and region, so a feature announced in October can reach your account days later. Paywalled pages aren’t supported as sources. Usage quotas are compute-based and unpublished, so treat the plans page as the source of truth rather than expecting a fixed daily allowance. And uploading documents to any AI tool raises the question of what happens to them — the AI chatbot privacy guide covers what to look for.

Start with one notebook of 5–10 sources on a topic you already know well — that way you can judge whether the answers deserve your trust before you rely on them for anything important. The open question is where Google takes the Gemini-app integration next: whether the standalone notebook remains the power-user home, or the Gemini app slowly absorbs it. Either way, the habit that matters is the same one the rename left untouched — put good sources in, and the answers have somewhere solid to come from.