NotebookLM is one of the few AI tools that people actually keep using, so we wanted to check in on how it is doing. The answer turned out to be more interesting than expected.

It is not called NotebookLM anymore. On July 16 Google retired the name and rebranded it Gemini Notebook, folding the product deeper into the Gemini family. Google has killed plenty of things this way, so if that made you nervous, fair enough. In this case the worry is misplaced, and that is the good news.

A month before the rename, the product quietly shipped the biggest upgrade in its history: a secure cloud computer attached to every notebook, native code execution, Gemini 3.5 as the default model, and agentic research that builds your source library for you.

Two more releases landed this month. And there is now genuine confusion about which product you are actually using, because Google split it across three surfaces and one of them still does not exist.

Here is what changed, what is worth your time, and what Google made needlessly confusing.

The June 8 release is the one that mattered, and it went out while the product was still called NotebookLM, which is part of why it got lost.

Every notebook now gets a secure cloud computing environment with native code execution. That is the headline. It means the tool stopped being a document summarizer and became something that can actually analyze your data, run calculations, generate charts, and produce spreadsheets from your sources rather than describing what is in them.

The model moved to Gemini 3.5 as default, with a transparent reasoning process that shows how a conclusion was reached rather than just asserting it. For anything where you have to defend the output, being able to follow the steps matters more than the answer being marginally better.

The third piece is the one that changes how you start a project. You can now open a chat about what you are researching, and the tool suggests sources using its own research skills and Google Search, building the knowledge base with you rather than waiting for you to upload files. Google specifically calls out finding primary sources in other languages and surfacing material from related authors. Sources only get imported with your explicit approval, which keeps the grounding intact.

That grounding is still the entire point. Gemini Notebook only knows what you give it, so answers come with citations back to your own documents rather than to whatever the model absorbed in training.

Two smaller releases landed in August, and the second is more useful than it sounds.

On August 7, Google added automated source updates through Workspace Studio for eligible Workspace editions, so a notebook built on live documents can refresh itself instead of going stale the moment someone edits the underlying file.

On August 17, it added full notebook copying, including eligible sources and Studio artifacts, which covers Audio and Video Overviews, Study Guides, Flashcards, Quizzes, and Slide Decks. Personal chat history and your own notes do not come along.

That is more significant than a convenience feature. It makes a notebook a template you can hand to someone else. Build a research setup once, with the sources and the generated study materials attached, and distribute it to a team or a class as a working artifact rather than a folder of files and instructions.

Here is where Google made this harder than it needed to be, and it is worth understanding before you go looking for a feature you read about.

Gemini Notebook is the standalone research app at notebook.google.com, with the full feature set. That is the real product.

Notebooks in the Gemini app is a different thing, launched back in April for organising chats and files. The notebooks sync between the two, but the feature sets do not match, so something you can do in one may be missing in the other.

A third surface, notebooks inside AI Mode in Search, was announced and has not arrived.

The version history is muddled too. The secure cloud computer, the model upgrade, and native code execution all shipped June 8, while the product was still called NotebookLM. The July 16 post announced only the rename. Google then updated the June post's timestamp, which is why coverage frequently reports both as a single release and why you will find articles claiming the cloud computer arrived with the rebrand. It did not.

If you rely on browser extensions or scripts that touch it, note that extensions needed updating and scripts need the new host added to their allowlist.

The interesting thing about Gemini Notebook is what it says about where Google thinks the value is.

Every other product in this space is racing toward more capability, bigger context, better benchmarks. Gemini Notebook competes on the opposite axis. Its entire pitch is constraint: it only knows what you give it, every claim traces back to a source you provided, and it will not wander into whatever the model half-remembers from training. In a year defined by hallucination rates and unverifiable vendor benchmarks, a tool that can only cite your own documents is a genuinely different product category.

Adding a cloud computer to that is what makes it more than a search box over your PDFs. You can now hand it a messy set of documents and get analysis, charts, and spreadsheets that trace back to specific pages, which is the actual shape of most knowledge work.

The practical advice is narrow. If you tried this a year ago and wrote it off as a summarizer with a podcast gimmick, it is a different product now, and the honest test is fifteen minutes with a real set of documents you know well enough to catch it being wrong. If it is already in your workflow, the notebook copying feature from last week is the one to look at, because turning a research setup into something shareable is the difference between a personal tool and a team one.

The stories moving fast beyond today's headlines:

The name is gone. NotebookLM became Gemini Notebook on July 16. The product did not shut down, it got absorbed into the Gemini family.

It got much better first. On June 8, still under the old name, it gained a secure cloud computer with native code execution, moved to Gemini 3.5, and started building your source library from a chat. That turned it from a summarizer into something that can analyze data and produce charts and spreadsheets grounded in your own documents.

Two more updates this month. August 7 brought automatic source refreshing for Workspace users. August 17 added full notebook copying, including the flashcards, quizzes, slide decks, and audio overviews, so a notebook becomes something you can hand to a team.

Google made it confusing. The real app is at notebook.google.com. The notebooks section inside the Gemini app is a different thing with different features. A third version for Search was announced and never shipped. Google also backdated the June post, which is why half the coverage says the cloud computer arrived with the rename. It did not.

Why it matters. Everything else in AI competes on knowing more. This competes on knowing less, only your documents, every claim traceable to a source you gave it. If you wrote it off as a summarizer with a podcast gimmick, it is a different product now.

Stay building. 🤖