Every general chatbot has the same flaw. Ask it about your own material, and it blends what you gave it with whatever it absorbed in training, and you cannot tell which is which. NotebookLM closes that gap by refusing to answer from anything except the sources you upload, and by citing the exact passage behind every claim. That single constraint is why researchers and students trust it with work they would not hand to ChatGPT. Google renamed it Gemini Notebook in July 2026, which has confused a lot of people, but nothing about how it works has changed.
What Is NotebookLM?
NotebookLM is a research tool that works only from material you supply. You create a notebook, add sources such as PDFs, Google Docs, web pages, YouTube videos, or pasted text, and then ask questions. Answers come with inline citations pointing at the specific passage used.
That grounding is the whole design. A general assistant draws on training data and the open web, so a confident wrong answer is always possible. NotebookLM restricts itself to your notebook, so when it is wrong you can usually see why by clicking the citation.
On 16 July 2026, Google renamed the product Gemini Notebook. It remains standalone, existing links still work, and the feature set carried over intact. Notebooks now also appear inside the Gemini app with syncing between the two, and Google has said Search integration is coming. Most articles online still use the old name, and so do most searches, which is why this review does too.
How Does NotebookLM Work?
The workflow has three parts. Add sources, chat with them, then generate outputs in the Studio panel.
Sources can come from your own files or from the web. Since late 2025, Deep Research can build a cited source list for you by searching the open web, which solves the blank-notebook problem that stopped a lot of people at the first step. Each source can hold up to 500,000 words, and local uploads are capped at 200 MB.
Chat behaves like a normal assistant except that it will tell you when your sources do not contain the answer. That refusal is a feature, not a fault, and it is the main reason the tool is trusted for academic and professional work.
Studio is where the product separates itself. From the same sources, it produces Audio Overviews, which are two-host podcast discussions of your material, plus Video Overviews in standard, vertical, and cinematic formats, slide decks, infographics, flashcards, quizzes, and data tables. The July 2026 update also gave each notebook a secure cloud environment that can write and run code, which rolled out to Ultra and eligible business accounts first.
Key Features
- Source-grounded answers. Every response cites the passage it came from, so verification takes seconds.
- Audio Overviews. Turn a stack of documents into a podcast-style conversation, now available in multiple languages.
- Video Overviews. Narrated visual summaries in standard, short-form vertical, and cinematic styles.
- Deep Research. NotebookLM finds and cites its own web sources, then adds them to your notebook.
- Study tools. Flashcards, quizzes, and study guides generated directly from your uploaded material.
- Mind maps and infographics. Visual breakdowns of how ideas in your sources connect.
- Slide decks and data tables. Exportable outputs, including PPTX, built from the same notebook.
- Code execution. Newer notebooks include a cloud environment that can write and run code on your data.
Free Plan Limits And Pricing
The free Standard tier includes every feature. Paid plans exist only to raise daily volume, which is unusual and worth stating plainly. Figures below reflect Google’s published limits as reported in August 2026.
| Limit | Free | Google AI Pro ($19.99/mo) |
|---|---|---|
| Notebooks | 100 | 500 |
| Sources per notebook | 50 | 300 |
| Chat queries per day | 50 | 500 |
| Audio Overviews per day | 3 | 20 |
| Video Overviews per day | 3 | 20 |
| Deep Research runs | 10 per month | 20 per day |
Google AI Plus at $4.99 per month sits between the two, and Ultra at $99.99 raises source capacity further. There is no standalone NotebookLM subscription any more. Older articles quoting “NotebookLM Plus at $19.99” are describing a plan structure that no longer exists, since the paid tiers now ride on Google’s AI subscriptions.
For most people, the free tier is genuinely enough. Fifty chats a day and three audio overviews cover normal study or research use. The caps that bite first are sources per notebook, if you are working through a large literature set, and Deep Research at ten runs a month.
Performance And Experience
Accuracy on uploaded material is the strongest thing about it. Because the model refuses to invent, the failure mode is usually an unhelpful answer rather than a confidently wrong one. That is the right trade for research work.
Speed varies by output. Chat is fast. Audio and video overviews run for minutes because they are producing media, and Deep Research takes longer again since it is searching and reading before it writes.
Ease of use is excellent. There is no prompt craft to learn, and adding a YouTube link or a PDF takes one click. The interface has stayed simple even as the feature list has grown.
The weak points are structural. Notebooks are silos, so you cannot search across all of them at once. There is no way to capture web pages as you read without switching tools. And newer features consistently land on expensive tiers first, which makes the free experience feel a step behind announcements.
Our Honest Verdict On NotebookLM
This is the rare Google product where the free tier is not a trap. Every feature is included, and payment only buys volume. For a student or an independent researcher, that is close to unbeatable value, and the citation-first design solves a real problem that general assistants have not. Audio Overviews still feel like a party trick until you use one to revise on a commute, at which point they become the reason you keep the tool.
The frustrations are real but narrow. Notebooks cannot talk to each other, which gets painful once you have thirty of them. The lack of a consumer API blocks anyone wanting to build on top of it, and people have been asking for years. The rename to Gemini Notebook was handled quietly enough that most of the internet still has the old information. And the best new features reach paying tiers first, so free users read about capabilities they cannot use yet. None of that outweighs what it does well. Start free, and only upgrade once you can name the limit you are hitting.
What NotebookLM Should Fix Next
- Add cross-notebook search so large collections stay usable.
- Release a consumer API, which has been among the most requested features for years.
- Provide a browser extension for capturing sources while reading.
- Publish a single, clearly dated limits page so users are not relying on third-party reporting.
- Make the rename obvious in-product, since many users still do not know it happened.
- Give free users a visible counter for remaining daily chats and generations.
