How to Use NotebookLM

NotebookLM turns the documents you already have into answers you can trace back to a source, then into briefings your team can use. Start with one business question, add the relevant contracts, research or SOPs, and check the citations before you act. The biggest payoff is less time reconstructing context before a client call, proposal or handover.

Google renamed NotebookLM Gemini Notebook in July 2026. It remains a standalone product. This guide uses NotebookLM because that is the name many people still search for; Google's current interface and help pages use Gemini Notebook. Google's naming announcement.

What Is NotebookLM?

NotebookLM is Google's AI research workspace for working with your sources. It combines cited answers with outputs such as briefings and overviews. Google's product guide.

Think of a notebook as a project binder with an assistant sitting beside it: the binder holds your evidence, chat helps you interrogate it, and Studio packages the findings. Keep the documents needed for one decision together: NotebookLM cannot query across multiple notebooks at once. Google's notebook guide.

For a consultant, that could mean a client binder containing the contract, discovery notes and market research. For an operator, it could be the approved procedures for a particular role. A well-chosen collection is more useful than a dump of every file your company owns.

My recommendation: start with a written brief. It is easier to inspect an answer beside its supporting passages than to spot a missing qualification while listening to a polished overview.

Free, Paid and Workspace Plans

Start with the free tier for a small personal pilot. For company documents, first check the access your employer already provides. Google's current plan table separates notebook capacity from AI usage. Google's plan comparison.

Capacities come from Google's help center; prices are the U.S. monthly prices on the Google AI plans page, checked October 14, 2026. These are broader Google AI subscriptions. Prices, capacity and availability can change; check your local offer before subscribing. Google also lists an enterprise licensing route through Google Cloud.

Workspace availability depends on the edition. Business Starter has standard access. Business Standard, Business Plus, Enterprise Standard and Enterprise Plus have higher access. Education Fundamentals and Standard have standard access; Education Plus and the Teaching and Learning add-on have more access; Google AI Pro for Education has higher access. AI Expanded Access and AI Ultra Access are additional upgrade routes. Check the badge beside your profile and ask your administrator about your assigned license. Google's Workspace edition table.

Do not treat an old daily chat allowance as universal. Google's general usage page now describes compute-based limits, meaning allowance depends on the work a request requires. It lists Plus at twice standard usage, Pro at four times standard, and Ultra at five or twenty times Pro, depending on the subscription. Quotas refresh every five hours until the weekly limit is reached. Open Settings → Usage for your account's allowance. Google's usage guide.

The Workspace help page separately continues to publish daily feature quotas. Use the documentation for your account type together with its Usage screen when planning a team rollout.

How to Use NotebookLM, Step by Step

1. Create a Notebook for One Decision

Open Google's notebook app in your browser with the intended account. Select Create new notebook, then Upload a source. For an existing notebook, use Add sources. Google's setup instructions.

Choose a title such as “Acme discovery brief” or “Support onboarding.” Keep unrelated clients in separate notebooks. Before uploading, decide what the finished answer must help someone do: prepare questions, identify missing evidence, explain a procedure or summarize decisions.

2. Add the Source Material

These are Google's supported inputs. A token is a small piece of text, rather than necessarily a whole word. Google's supported sources.

The caps are 500,000 words per source and 200 MB per uploaded file. Web imports omit embedded media and nested pages; paywalls are unsupported. Google-file comments and footnotes are omitted. Drive imports auto-sync, with a manual sync option.

Name your files by subject and date. For a contract, include amendments alongside the original. For SOPs, choose the approved version. For research, record when a price or claim was captured. These small habits make contradictions easier to investigate later.

3. Ask a Question That Defines the Job

Select the sources you want to use, then ask in chat. A strong prompt specifies the audience, the decision, the desired structure and what to do when evidence is missing. Google supports choosing the sources used for an answer. Google's chat guide.

Try this:

Using the selected sources, prepare a briefing for the project lead. Identify the agreed scope, deadlines, dependencies and unresolved questions. Cite each factual point. Separate explicit statements from your inferences. If an answer is missing or sources conflict, say so and identify what a person should check.

A prompt is a request, not an accuracy guarantee. Ask a follow-up about a narrow point if the first answer bundles several different obligations together.

4. Open the Citations

Hover over a citation to see the quoted text, or select it to open the supporting passage in context. Use Save to note to retain a useful response with its clickable citations. Google's citation instructions.

Check the claims that change the decision: the renewal date, exception, deliverable, owner or customer promise. Read the surrounding paragraph. A citation can point to a real passage while the answer still overstates what it means.

For contracts, use this as preparation for the responsible reviewer. Finding the clause and deciding its legal effect are different jobs.

5. Choose a Studio Output

Match the output to how the recipient will use it. Google lists all the following in its Studio guide.

For a first briefing, open Studio → Reports → Document, choose a suitable format and customize the instructions. Ask for the audience, key decisions and unanswered questions you need. Review the result before circulation. Google's report instructions.

Reports can export to Docs, and Data Tables can export to Sheets. Changes in those exported files do not sync back, and notebook sharing permissions do not carry over. Check the destination file before sending it. Google's export notes.

Use the browser for this walkthrough. Availability varies by feature, age and account: Google currently limits Infographics and Slide Decks to adults, and Mind Maps are not supported in the mobile app. Cinematic Video Overviews are adult-only and English-only. The linked feature pages give the applicable conditions.

Five Business Workflows Worth Trying

These are proposed workflows, ranked by likely business value. The prompts ask NotebookLM to prepare material for review; they do not imply automatic approval, submission or task execution.

1. Prepare an RFP Response with Evidence

Who: a sales engineer or proposal consultant whose subject experts keep answering the same questions. An RFP is a request for proposal: the buyer's requirements and questions for potential suppliers.

Add: the RFP and amendments, an approved answer library, current product documentation, security policies and relevant case studies. Exclude superseded promises unless you need to compare versions.

Prompt:

Create an RFP preparation table using only the selected sources. For each requirement, include its identifier, a proposed answer, supporting citations and evidence status: supported, conflicting or missing. Distinguish the buyer's request from what our documents actually promise. Do not invent certifications, service levels, pricing or roadmap commitments. Finish with questions for the proposal owner.

Payoff: experts can spend their review time on gaps and commitments. The proposal owner still checks completeness against the original RFP and approves every external claim.

2. Build a Client Research Pack

Who: a consultant preparing for discovery or a founder pursuing a complex account.

Add: discovery notes, the prospect's public strategy and product pages, relevant research, the signed scope if one exists, and prior meeting transcripts. Include source dates in filenames.

Prompt:

Prepare a discovery pack for a consultant meeting this client. Summarize the business model, stated priorities, known constraints and unresolved questions. Cite factual claims. Separate client statements, external research and your hypotheses. Suggest questions that would test the hypotheses. Highlight where a proposed project might exceed the signed scope, without making a legal determination.

Payoff: a stronger call agenda and fewer repeated questions. Treat hypotheses as questions to investigate, not facts to repeat in a pitch.

3. Onboard a New Hire from SOPs

Who: an operations lead whose experienced staff repeatedly explain the same procedures. SOPs are standard operating procedures: the documented steps for doing a job.

Add: approved SOPs, the role description, escalation rules, product references and examples of completed work. Leave out draft policies and unrelated employee records.

Prompt:

Create a first-week learning plan for a new customer operations hire using these approved sources. Organize it by tasks they need to perform, prerequisites and escalation points. Cite the procedure for each task. Flag conflicting instructions. Then propose realistic practice scenarios and identify which source supports the correct action. Do not invent company policy.

Payoff: the manager can review one coherent learning path, then use Studio Quizzes or Flashcards for practice. Keep the original SOP as the operating authority and give the hire a named person for unresolved questions.

4. Turn Meeting Notes into a Decision Digest

Who: a project lead inheriting several weeks of meetings.

Add: dated meeting notes or transcripts, the current project plan, previous decision logs and the existing action list. If using audio, use a supported recording you are authorized to process.

Prompt:

Create a chronological decision digest from these meetings. Separate decisions, proposals, actions, blockers and unanswered questions. For actions, include the owner and due date only when explicitly stated; otherwise write “not specified.” Cite the source for each entry. Identify later decisions that appear to change earlier ones and preserve both dates for review.

Payoff: the next meeting can focus on unresolved work. Compare the digest with your actual task tracker before updating owners or deadlines.

5. Prepare a Competitor Brief

Who: a founder or product marketer preparing positioning for a specific deal or segment.

Add: dated competitor pricing and product pages, release notes, public case studies, authorized sales notes and your own current product documentation.

Prompt:

Compare these competitors for a buyer choosing a customer-support platform. Build a table of target customer, documented capabilities, pricing conditions and stated limitations. Cite every factual comparison and show the date where available. Separate vendor claims from independent evidence. Mark missing information as unknown. Finish with buyer questions that could change the decision.

Payoff: a brief the sales team can challenge and update. A missing feature in your collection does not prove the competitor lacks it.

The Business Math: Measure the Review Time Too

The budget line to test is research and preparation labor. A free account does not make the whole workflow free: someone must choose sources, verify claims and keep the material current.

Here is an illustrative planning calculation, not a measured NotebookLM result. Suppose four monthly briefs currently take three hours each, and internal time costs $80 an hour. That is $960 per month. If a pilot brings the entire process, including review, down to one hour per brief, labor becomes $320. The difference is $640 minus any incremental subscription cost. Add initial setup time separately.

Replace those assumptions with your own timed pilot. If collecting documents and checking the draft consumes the same time as the original job, the workflow has not yet earned an upgrade. Buy more capacity when a measured recurring job needs it.

Two Services You Could Build Around This Workflow

Strongest Opportunity: RFP Evidence Preparation

A proposal specialist could sell a reviewed evidence pack to small B2B vendors: organized source material, a requirement table, draft answers and an exception list for the customer's experts.

DataForSEO estimates 1,300 monthly U.S. searches for “AI RFP response software”, checked October 14, 2026. That measures searches, not buyers, but the query expresses a specific commercial need. A visible price anchor already exists: DeepRFP lists $89 per user per month, billed monthly.

The smallest sellable version could be one manually managed bid, with the buyer's approved documents, a repeatable prompt set and a signed-off handover. The hard part is keeping evidence current and getting accountable experts to approve commitments. NotebookLM helps prepare the material; it does not make the service's professional judgment defensible by itself.

This is the strongest opportunity because the job has a deadline, an identifiable owner and expensive interruptions. The value is the reviewed evidence and gap resolution you deliver.

Second Opportunity: Role-Specific Onboarding Packs

An operations consultant could package a company's SOPs into a role-specific learning plan, scenario quiz and manager review checklist.

DataForSEO estimates 480 monthly U.S. searches for “employee onboarding checklist” in the same check. This is broader informational demand, so it is weaker evidence of willingness to pay than the RFP query.

A first version could cover one role with one source owner and a review date. Deliver the notebook, checked learning material and a process for handling policy changes. The catch is source quality: contradictory SOPs become contradictory training unless someone resolves them. A quiz result also does not establish that a new hire can perform the work.

Both opportunities can begin as services using the documented interface. Test demand with a concrete deliverable before designing integrations or a larger software product.

Workspace Data and Privacy

For confidential business material, check the exact license and your organization's rules before choosing the account. Google's Workspace table distinguishes core-service access with enterprise protections from additional-service access, including Workspace Individual. For qualifying core-service editions, uploaded files, chats and outputs are not reviewed by humans or used to improve generative AI models. Google's account and data-handling table.

Google's Workspace Privacy Hub adds important boundaries:

• Core-service use falls under the organization's Workspace agreement, including the Cloud Data Processing Addendum.
• Importing a Drive source creates a separate copy stored with Notebook data.
• Organizational Drive file-sharing and data-region settings do not apply to that Notebook data.
• Workspace Data Loss Prevention, or DLP, is not currently integrated with Notebook. Administrators can control who has access to the service.

These are Google's published product boundaries. For client contracts and internal policies, have the account owner confirm that this storage and sharing arrangement fits the work. Google Workspace Privacy Hub.

Where NotebookLM Stops Helping

NotebookLM cannot resolve a contradiction your business has never decided. It can help expose the conflicting statements, but a person must decide which policy, promise or document governs the work.

Nor does a polished artifact establish accuracy. Google explicitly cautions that reports, audio and visual outputs can contain mistakes. Keep a written, checked version of the claims that matter before producing a presentation or narrated briefing. Report accuracy guidance, audio guidance.

For a different research or knowledge-management workflow, use our NotebookLM alternatives comparison to evaluate the options against your actual job.

How to use a Gemini notebook?

Open the notebook app, create a notebook, add the documents for your project and ask a focused question. Inspect the citations, then choose a Studio output for the audience. Gemini Notebook is the current name for NotebookLM.

Can NotebookLM interpret images?

Yes. Google's supported-source list includes image uploads, and its chat guide explains that citations can refer to images as well as text. Inspect important details yourself, especially small labels and ambiguous visuals. Google's chat guidance.

What are the limits for the free version of NotebookLM?

Google's current plan table gives Standard users 100 notebooks and 50 sources per notebook. AI usage is a separate allowance. Check Settings → Usage instead of relying on a fixed daily quota quoted in an older tutorial. Google's plan table, usage guidance.

Can NotebookLM hallucinate?

Yes. It can generate inaccurate material. Citations make the supporting evidence easier to inspect; they do not remove the need to check whether the answer follows from it. Google warns that generated reports may contain errors. Google's report guidance.

Your Monday move: choose one upcoming client call or RFP, gather its approved documents, and run the matching prompt above. Have the responsible person review the brief and record the total time. Repeat on the next comparable job before expanding the workflow.

If you want a repeatable document-to-briefing workflow built for your team, explore AI automation.