How to Use NotebookLM (Now Gemini Notebook) for Work: A Practical Guide
NotebookLM is now Gemini Notebook: what changed
On 16 July 2026, Google renamed NotebookLM to Gemini Notebook. Google described it as the same standalone research tool, now connected more closely to the rest of Gemini. Three things changed in practice:
- The name. Most guides and videos online still say NotebookLM. They describe the same product.
- Syncing with the Gemini app. Notebooks sync between standalone Gemini Notebook and the Gemini app, and Google says Search’s AI Mode is next.
- Code execution. Notebooks can now run code in a secure cloud environment for data analysis. When this guide was written, that feature had reached Google AI Ultra users and some Workspace customers, with Pro users on the web getting it gradually.
What didn’t change is the most important part: answers come from the sources you add, each with a citation you can click to see the passage it came from. Google says more than 30 million people and over 600,000 organisations use it.
Why it is different from ordinary chat
Ask a general chatbot about your leave policy and it will guess, using what it knows about leave policies in general. Ask Gemini Notebook, after adding your actual policy, and it answers from your policy and shows you the clause.
That makes it suited to work where being wrong is expensive and checking matters: HR and compliance questions, research, onboarding, proposal preparation and meeting follow-up. It is less suited to open-ended creative writing, where a general assistant is the better tool.
How to set up your first notebook in ten minutes
- Open Gemini Notebook with your Google account, either the standalone tool or from inside the Gemini app. At work, use your company Google Workspace account if your organisation has enabled it, rather than a personal Gmail account.
- Create a notebook for one purpose only, for example “Travel and expense policy 2026”, not “HR stuff”. Narrow notebooks give sharper answers.
- Add your sources: files such as PDFs and documents, Google Docs and Slides, web links, YouTube videos and pasted text.
- Read the automatic summary it generates, and check it picked up the right documents.
- Ask a question you already know the answer to. Click the citation and check it points to the right passage. This one test tells you whether your sources are clean enough to trust.
A naming habit that saves time later: start every notebook name with the team or project and the date, for example “Sales | Acme renewal | Oct 2026”. Once a team has dozens of notebooks, this matters.
Plan limits you should know about
Limits depend on your plan. These are Google’s published figures at the time of writing. Google marks them as subject to change, so check its help page before you plan a big project around them.
| Plan | Notebooks per user | Sources per notebook |
|---|---|---|
| Standard (free) | 100 | 50 |
| Plus | 200 | 100 |
| Pro | 500 | 300 |
| Ultra | 500 | 500–600 |
Enterprise options are also available through Google Cloud. For most teams, the source limit matters more than the notebook limit: a notebook holding a whole department’s archive is usually a sign it should be split.
How to ask questions that return useful, cited answers
Most disappointing answers come from vague questions. These four patterns work consistently in the sessions I run:
“What is the approval limit for client entertainment? Quote the exact clause and tell me which document it comes from.”
“Compare how the 2024 and 2026 versions of the policy treat working from abroad. Put the differences in a table.”
“Based only on these sources, what questions would a new joiner have that the documents don’t answer?”
“You are briefing a sales manager who has five minutes. Summarise the three findings from these reports that matter most for pricing, with citations.”
If an answer has no citation, or the citation doesn’t support the claim, treat the answer as unreliable. The same discipline applies to any AI tool. For more general patterns, the site’s Gemini prompts collection is a good companion, written for students but just as useful at work.
Studio outputs: audio, video, slides, mind maps and reports
The Studio panel turns your sources into other formats. Google’s documentation lists reports, flashcards, infographics, slide decks, Audio Overviews and Video Overviews, and mind maps let you explore how ideas in the sources connect. Here is where each one helps at work:
- Audio Overview: a podcast-style conversation about your sources. Good for a long policy pack or a research bundle you want people to absorb on the commute.
- Video Overview: a narrated visual summary, useful for onboarding and internal training.
- Slide deck: a first draft of a presentation built from the sources. Treat it as a starting structure and edit before presenting.
- Reports and briefing documents: a structured written summary, such as a study guide, briefing doc or FAQ.
- Mind map: a quick way to see what a large set of sources covers before you start asking questions.
- Flashcards and quizzes: surprisingly useful for compliance refreshers and new-joiner training.
Studio outputs are only as accurate as the sources, and they can still oversimplify. Check anything that will be shared outside your team.
Rolling this out to a whole team?
Workflow 1: an HR policy and onboarding assistant
Sources: the employee handbook, leave and travel policies, benefits summary, IT acceptable-use policy and the onboarding checklist.
How to use it: HR builds and owns the notebook, then shares it with new joiners as view-only. New joiners ask “How do I claim my internet allowance?” and get the answer with the clause attached. HR can generate an Audio Overview as a welcome briefing and flashcards for the policy quiz.
Watch out for: stale versions. When a policy changes, replace the source the same day, or the notebook will confidently quote the old rule. A short rule on approved tools and data, like an AI usage policy for employees, is a useful first source.
Workflow 2: turning meeting transcripts into decisions
Sources: transcripts or notes from a series of meetings, such as a month of project calls.
Ask: “List every decision made, who made it and the date, with a citation.” Then: “List open action items with owners. Flag any that were raised more than twice without being closed.”
This is where a notebook beats a single meeting summary: it sees across meetings, so it can spot the item that keeps slipping. Get attendees’ agreement before recording, and don’t upload transcripts of confidential HR or legal discussions.
Workflow 3: sales and proposal research
Sources: the client’s annual report, recent press releases, their careers page, your previous proposals to similar clients, and your own case studies.
Ask: “What are this company’s three stated priorities for the year, with citations? Which of our case studies is most relevant to each?” Then generate a briefing document for the account team before the meeting.
Keep one notebook per account, and delete it or archive the sources when the deal closes.
Workflow 4: market and competitor research
Sources: analyst reports you have licensed, competitors’ public web pages, product documentation and relevant YouTube talks.
Ask: “Build a comparison table of these five competitors on pricing model, target customer and main claims. Cite each cell.” Follow up with: “Where do these sources disagree with each other?” Disagreements are often the most useful finding.
With code execution, analysts on supported plans can also run simple analysis on data files inside the notebook. Check the numbers before they go into a deck.
Workflow 5: training and learning content
Sources: a course syllabus, internal process documents and recorded training sessions.
Use it to: generate a study guide, quiz questions and flashcards, then an Audio Overview people can replay. Trainers can ask “Which concepts in this material are most likely to confuse a beginner?” to plan their sessions.
Workflow 6: legal and contract review support
Sources: a contract and the related templates, or a set of regulations.
Ask: “List every obligation on the supplier, with clause numbers.” “Where does this draft differ from our standard template?”
It is a reading assistant, not legal advice. Lawyers still interpret. They just get to the right clauses faster. If your legal team drafts with AI as well, these AI prompts for legal drafting pair well with a contract notebook.
Sharing notebooks with your team
Notebooks can be shared, and how much control you have depends on your account type. In practice:
- One owner per notebook. Someone has to be responsible for keeping sources current.
- View-only for most people. Readers ask questions, owners change sources.
- Use Workspace accounts at work. They come with your organisation’s admin controls and terms. Personal accounts don’t.
- Name and archive. Consistent names, and a quarterly clean-up of notebooks nobody has opened.
Data privacy: what not to upload
Google says Gemini Notebook does not use the content you upload to train its models, and Workspace accounts come with additional enterprise data protections. Even so, “not used for training” is not the same as “allowed by your company”. A sensible default list of things to keep out:
- Personal data of customers or employees, unless your organisation has explicitly approved it. In India the Digital Personal Data Protection Act applies, and in the UAE usually the federal PDPL or the DIFC and ADGM regimes.
- Client documents covered by a confidentiality agreement that doesn’t allow third-party processing.
- Unreleased financial results and price-sensitive information.
- Passwords, keys and access credentials, which should never be in any document.
- Anything your company classifies as restricted.
When in doubt, check Google’s current privacy terms for your account type and ask your IT or data protection team before uploading.
Gemini Notebook vs ChatGPT Projects vs Claude Projects
All three let you work with a set of your own files, and they suit different jobs:
| Gemini Notebook | ChatGPT Projects | Claude Projects | |
|---|---|---|---|
| Best at | Answers grounded in your sources, with citations | General work around a shared set of files and chats | Long documents, careful writing and analysis |
| Stays within your sources? | Yes, by design | Not strictly; it also draws on general knowledge | Not strictly; it also draws on general knowledge |
| Turns sources into | Audio, video, slides, mind maps, reports, flashcards | Documents and other files through chat | Documents and artifacts through chat |
| Natural fit if you use | Google Workspace | ChatGPT Business or Enterprise | Claude Team or Enterprise |
Many teams use more than one: Gemini Notebook when the answer must come from the documents, and a general assistant for drafting what comes next. If you are weighing the general assistants against each other, this Claude vs ChatGPT comparison covers the trade-offs.
A 30-minute exercise to get your team started
Run this in a team meeting. It works better than any demo.
- Minutes 0–5: everyone creates a notebook and adds two or three documents they use every week, after checking the “what not to upload” list.
- Minutes 5–15: each person asks three questions they already know the answers to, and checks every citation.
- Minutes 15–25: each person tries one comparison question and one “what is missing” question, then generates one Studio output.
- Minutes 25–30: go round the table. Which answer was most useful? Which one was wrong, and why? Agree on one workflow the team will try for two weeks.
The “which answer was wrong” conversation is the most valuable part. It teaches people to check citations, which is the habit that makes this tool safe to rely on. If you want to see how much time the new workflow actually saves, this guide on how much time AI really saves shows how to measure it honestly.
Questions people ask
Want your team using Gemini Notebook on real work?
About the author
Hitesh Motwani is a generative AI trainer who runs hands-on workshops for teams in India and the UAE, delivered in the UAE through Skillopedia FZC LLC. He is visiting faculty at K J Somaiya Institute of Management and the author of Generative AI 360° (ZebraLearn, 2025).