Gemini Notebook is the tool formerly known as NotebookLM: you upload your own documents, and Google's models answer questions using only those documents, with citations back to the exact passage. This review covers what the rename changed, which sources it accepts, how well the grounded answers hold up, what the Studio outputs (podcasts, videos, mind maps, slides) are good for, the new code execution feature, and the free versus paid limits that actually affect daily use. Prices and limits are as of July 2026.
Verdict: Gemini Notebook is the best tool available for asking questions of a fixed pile of documents, and the free tier is generous enough for most students and casual researchers. Pay for Google AI Pro if you work with more than 50 sources per project or need more than a handful of Audio Overviews a day. Skip it if you want a general-purpose chatbot or a place to write; it is a reading tool, not a writing tool.
What Gemini Notebook is (and why the name changed)
Google renamed NotebookLM to Gemini Notebook on July 16, 2026. The product itself did not move: it lives at the same address, your existing notebooks and shared links still work, and Google describes it as remaining "a standalone product focused on being your premier research tool." The new name mostly signals that it is now part of the Gemini family, with a blue-violet Gemini gradient on the logo and notebooks that sync with the Gemini app.
The rename came with two substantive additions, which 9to5Google's coverage lays out: every notebook now has access to a secure cloud computer that can write and run code for data analysis, and notebooks are coming to AI Mode in Google Search. Google also said more than 30 million people and over 600,000 organizations use the product, which explains why it got promoted from side project to first-class brand.
The core idea has not changed since launch. Unlike a normal chatbot, Gemini Notebook answers from your sources, not from whatever the model absorbed in training. That single design decision is what makes it useful for research, and it is the lens for the rest of this review.
Setting up a notebook: sources it accepts
A notebook starts empty. You add sources, and the model reads only those. The list of accepted formats has grown a lot over the past year. According to DigitalOcean's overview, a notebook can take:
- Documents: PDF, DOCX, TXT, CSV, Google Docs, Google Sheets and Google Slides
- Media: MP3 and WAV audio, MP4 video, and PNG, JPG, GIF or WEBP images (with OCR for text in images)
- Web content: direct URLs, YouTube videos and results from a built-in web search
- Pasted text and files from Google Drive
Each source can be up to roughly 500,000 words, and the whole notebook is read through a context window of up to one million tokens. In practice, that means you can drop in a 300-page technical manual, an hour of recorded lecture audio and a handful of web pages and ask questions across all of them in a single chat.
Two setup features are worth knowing. The "Discover sources" option lets you describe a topic and have the tool go find web pages for you, and Deep Research, added in November 2025, goes further by compiling a citation-backed report that you can then add as a source. Both are useful for building a starting notebook quickly, with the obvious caveat that you should skim what it found before trusting answers built on it.
Chat that sticks to your documents
This is the part that has to work, and it mostly does. Ask a question, and the answer comes back with numbered citations; click one and the source panel jumps to the highlighted passage. When the answer is not in your sources, it usually says so rather than improvising, which is the behavior you want and which general chatbots still get wrong often enough to matter. If you have been bitten by that before, our guide to checking and fixing AI hallucinations explains why grounding helps.
The quality of the reasoning depends on the model behind it. Free users get Gemini's Flash-class model, which is fast and fine for lookup-style questions. Gemini 3.1 Pro, released February 19, 2026, is available in Gemini Notebook only to Google AI Pro and Ultra subscribers. The difference shows on synthesis questions ("what do these five papers disagree about?") rather than retrieval questions ("what does the contract say about termination?").
Chat history is saved, so you can close the tab and pick up a conversation later, and you can set a custom goal for the notebook of up to 5,000 characters, which acts like a standing instruction: "answer as a tax adviser," "always quote page numbers," and so on. Both arrived in October 2025 and make long projects far less annoying than the original version, which forgot everything on refresh.
Where grounded chat falls down is nuance across very long sources. Ask something whose answer is scattered across a 400-page PDF and you will sometimes get a partial answer with a confident tone. The citations let you catch this, but you have to actually click them.
Audio Overviews, Video Overviews, mind maps and slides
The Studio panel on the right of each notebook is where the product turns from useful into fun. With one click you can generate, from your sources:
Scroll sideways to see the full table →
| Output | What you get | Best for |
|---|---|---|
| Audio Overview | A two-host podcast-style conversation about your sources | Absorbing material on a commute |
| Video Overview | A narrated video with generated visuals | Quick explainers to share |
| Mind Map | A branching diagram of concepts, clickable to drill in | Getting the shape of an unfamiliar topic |
| Slide Deck | An editable presentation, exportable as PPTX | First drafts of talks |
| Infographic | A single-page visual summary | Sharing takeaways |
| Data Table | Structured comparison table, exportable to Sheets | Turning prose into rows and columns |
| Report | A written summary in a chosen format | Briefing documents |
| Quiz and Flashcards | Self-test material | Studying |
Audio Overviews remain the signature feature. The hosts are engaging, and you can steer the episode with a focus prompt before generating. The failure mode is that they can be relentlessly upbeat about dull material and occasionally simplify to the point of being slightly wrong, so treat them as an on-ramp, not a substitute for the reading.
Slide Decks improved a lot in February 2026 when Google added prompt-based editing ("make slide 5 more concise") and PPTX export. The output is still a first draft, but it is a first draft with the right facts in it, which is more than most slide generators manage.
Video Overviews come in two flavors. The standard version is available broadly. The cinematic version, which uses Google's Veo 3 model to generate footage, arrived in March 2026 and is limited to Google AI Ultra subscribers, English only, with a small daily cap. It looks impressive in demos; for the price, most people will never use it.
The new code execution for data analysis
The headline of the July rename is that notebooks can now run code in what Google calls a secure cloud computer. Upload a CSV or a spreadsheet, ask for a breakdown by region or a chart of the trend, and the model writes and executes the code rather than guessing at arithmetic from the text.
The rollout is staggered. Google's announcement says code execution is available first to Google AI Ultra users and certain Workspace customers, with all Pro users to follow. In the early build, it is genuinely useful for the kind of light analysis that used to require exporting to Sheets: sums, group-bys, simple regressions and charts. It is not a replacement for a proper notebook environment, and the sandbox has no internet access, so it cannot fetch fresh data. But the combination of "answer from my documents" and "do the math correctly" closes the biggest gap the product had.
Free vs Google AI Pro: limits that actually matter
Gemini Notebook does not have its own subscription. It is bundled with Google's AI plans, which were restructured at I/O in May 2026 into Free, Plus, Pro and Ultra. The quotas that matter, from Google's own free-tier FAQ and DigitalOcean's plan breakdown, with Ultra shown as a range because its two tiers carry different caps:
Scroll sideways to see the full table →
| Limit | Free | Google AI Plus | Google AI Pro | Google AI Ultra (two tiers) |
|---|---|---|---|---|
| Notebooks | 100 | 500 | 500 | 500 |
| Sources per notebook | 50 | 100 | 300 | 500 to 600 |
| Chat queries per day | 50 | 100 | 500 | 2,500 to 5,000 |
| Audio/Video Overviews per day | 3 | about 6 | 20 | 100 to 200 |
| Gemini 3.1 Pro in Notebook | No | No | Yes | Yes |
| Code execution | No | No | Rolling out | Yes |
Google AI Pro costs $19.99 a month and also includes 5 TB of storage (up from 2 TB since April 1, 2026), the full Gemini model lineup in the Gemini app, Deep Research, YouTube Premium Lite and Workspace integrations. If you are deciding between Google's plan and its rivals, our comparison of the $20 AI subscriptions weighs the whole bundle.
For Gemini Notebook specifically, the limits that bite are sources per notebook and Audio Overviews per day. Fifty sources is plenty for a course or a single project; it is not enough for a literature review with 80 papers. Three Audio Overviews a day sounds fine until you start regenerating with different focus prompts. Fifty chat queries a day is more than it sounds, because each query can be long, but a heavy afternoon of research can hit it.
What it does badly
- It is not a writing tool. You can generate reports and notes, but there is no proper editor, no version history and no real way to build a document over weeks inside the product. Export and finish elsewhere.
- Source management is basic. There are no folders inside a notebook, no tagging and no way to search across notebooks. Once you have 40 sources, finding one is a scroll.
- Sources are snapshots. A web page or Google Doc is captured at the time you add it. Google Docs can be re-synced manually, but nothing updates automatically, which surprises people who expect a live link.
- Long-source answers can be partial. As noted above, the model sometimes stops at the first relevant passage in a very long document. Ask follow-ups.
- Audio Overviews are English-first and sometimes glib. Other languages exist but the hosts are noticeably less natural.
- Feature availability is confusing. Between Free, Plus, Pro, Ultra, Workspace and staggered rollouts, it is genuinely hard to know which features you have until you click.
Verdict: who should use Gemini Notebook
Pros
- Answers grounded in your own sources, with citations you can check
- Wide format support including audio, video, images and YouTube
- Audio Overviews and Mind Maps are the best of their kind
- Generous free tier
- Code execution finally makes numeric questions trustworthy
Cons
- Weak for writing and long-term document building
- Sources are static snapshots
- Best model and code execution are paid, and rollout is uneven
- No cross-notebook search
Who should use it: students, journalists, lawyers, analysts, anyone whose job involves understanding a fixed set of documents faster. The free tier covers a surprising amount of this.
Who should pay for Pro: people running several large projects at once, anyone regularly above 50 sources per notebook, and anyone who wants Gemini 3.1 Pro for synthesis-heavy questions.
Who should skip it: people who want a general assistant that also browses the web, writes emails and edits code. That is what the Gemini app, ChatGPT and Claude are for; Gemini Notebook is deliberately narrower.
Start on the free plan, hit the 50-source or three-podcast limit, and you will know whether $19.99 a month is worth it for you.
Sources and further reading
- NotebookLM is now Gemini Notebook (Google Blog)blog.google
- Google renames NotebookLM to Gemini Notebook (9to5Google)9to5google.com
- What Is NotebookLM? Features and How to Use It in 2026 (DigitalOcean)digitalocean.com
- NotebookLM Update 2026: Every New Feature Explained (Fello AI)felloai.com
- Frequently asked questions (Gemini Notebook Help)support.google.com
- Gemini 3.1 Pro: A smarter model for your most complex tasks (Google Blog)blog.google
- Google's new AI subscription plans explained (TechCabal)techcabal.com