Every other AI tool has the same problem: you ask a question and it answers from its training data. Sometimes those answers are right. Sometimes they're confidently wrong. You can't always tell the difference.
NotebookLM only answers from documents you upload. If it's not in your sources, it won't make something up.
That single constraint makes it the safest AI tool available for anyone who works with documents — which is everyone reading this.
I use it weekly. Board meeting prep. Policy analysis. Course development. Converting hundreds of hours of professional lectures into structured courseware for a client. It's become one of the tools I recommend first in my AI Leadership course, because it requires zero AI experience and the risk of hallucination is close to zero.
Here's the complete guide to using it well in 2026.
Your First Notebook: A Detailed Walkthrough
If you've never used NotebookLM — or never used any AI tool — this section walks you through everything. Each step describes exactly what you'll see on screen.
Step 1. Open your browser (Chrome, Firefox, Edge — any modern browser works) and go to notebooklm.google.com.
Step 2. You'll be asked to sign in with a Google account. Any Google account works — personal Gmail, Google Workspace, school-issued. If you're already signed into Google, it may skip this step entirely.
Step 3. You'll land on a dashboard. It's mostly empty if this is your first time. In the center of the screen you'll see a large "Create new" button (it may also appear as a "+" icon or "New Notebook" depending on your screen size). Click it.
Step 4. A blank notebook opens. At the top-left corner, you'll see placeholder text that says something like "Untitled notebook." Click that text. A cursor appears. Type a name that reflects your project — something like "Q2 Board Prep" or "Policy Review 2026." Press Enter.
Step 5. The left side of the screen shows a "Sources" panel. This is where your documents live. You'll see a prompt asking you to add sources, with several options:
- Upload — click this to browse your computer for files. It accepts PDFs, Word documents (.docx), PowerPoint files (.pptx), plain text, images, audio (MP3, WAV), and video (MP4). Maximum file size is 200MB.
- Google Drive — click this to connect to your Drive. You can pull in Google Docs, Sheets, and Slides directly without downloading them first.
- Website — paste any URL. NotebookLM reads the page and adds its content as a source. Useful for blog posts, research articles, or documentation pages.
- YouTube — paste a YouTube video link. The video must have captions (most do). NotebookLM ingests the transcript.
Step 6. Start by uploading two or three documents. Click "Upload," navigate to the files on your computer, and select them. You can select multiple files at once. They'll upload in a few seconds and appear as cards in the Sources panel on the left.
Step 7. Once your sources are loaded, NotebookLM automatically generates a summary and a set of suggested questions in the main area of the screen. Read these — they're surprisingly useful for getting oriented. The summary captures the main themes across all your documents, and the suggested questions point you toward the most interesting connections.
Step 8. At the bottom of the screen, you'll see a text box that looks like any chat interface. Type a question — something specific to your documents. For example: "What are the three main recommendations in the strategic plan?" or "Does the policy handbook mention anything about personal device use?" Press Enter.
Step 9. The response appears in the main panel. Here's where NotebookLM earns its trust: you'll see small numbered citations scattered through the text, like footnotes. Hover over any number and a tooltip shows the exact quote from your document. Click the number and the Sources panel scrolls to that passage, highlighted. You can verify any claim in seconds without hunting through pages yourself.
That's a working notebook. No API keys, no configuration files, no prompt engineering formulas. Upload your documents, ask your questions, verify with citations.
Audio Overviews: The Feature Everyone Talks About
Audio Overviews generate a podcast-style conversation where two AI hosts discuss your uploaded documents. It's the feature that made NotebookLM go viral, and it's genuinely useful — not just a novelty. Here's how to use it.
Step 1. With your notebook open and sources added, look at the right side of the screen (or the top if your window is narrow). You'll see a panel labeled "Studio." Click it to expand the studio options.
Step 2. Click "Audio Overview." A dialog may appear letting you optionally guide the focus (e.g., "Focus on the budget section" or "Explain the policy changes"). You can leave this blank for a general overview, or type a focus area if you want something specific.
Step 3. Click "Generate." NotebookLM takes one to three minutes to produce the audio, depending on how much source material you've loaded. You'll see a progress indicator.
Step 4. When it's ready, a player appears. Press play. Two AI hosts have a natural-sounding conversation about the key points in your documents — they summarize, explain, compare sections, and highlight what's interesting. It runs 10-20 minutes depending on the volume of your sources.
Step 5. Here's the part most people miss: Interactive Mode. While the audio is playing, look for a "Join" button or a hand-raise icon near the audio player. Click it.
Step 6. Type your question (or use your microphone if prompted). The AI hosts pause their conversation, address your question using only your source documents, and then pick back up where they left off. It's like being able to interrupt a podcast to ask the hosts a follow-up — and having them actually answer from the material.
I use Audio Overviews for two things: reviewing long documents during my commute, and creating briefings for stakeholders who won't read a 30-page report but will listen to a 15-minute summary.
The Features That Actually Matter
NotebookLM has added a lot of features over the past year. Beyond Audio Overviews and citations, here are the ones I use.
Data Tables let you extract and compare specific information across multiple documents. Upload three vendor proposals and ask it to build a comparison table of pricing, features, and contract terms. The table exports to Google Sheets. I used this when converting a client's lecture content — pulling question banks from multiple source files into structured comparison tables saved hours of manual work.
Mind Maps generate visual topic diagrams from your sources. Useful for getting a quick structural overview of a complex document before you start asking detailed questions.
Quizzes and Flashcards are built for educators, but useful for anyone who needs to test their understanding of a document set. The quizzes have an "Explain" button that shows you the source passage behind each answer.
Slide Decks generate presentations from your sources with one click. Since February 2026, you can edit them with prompts and export to PPTX. Not production-quality, but a fast starting point.
Deep Research runs multi-step web research and adds the results as a new source in your notebook. This is useful when your uploaded documents reference external data you want to verify or expand on.
Five Ways I Actually Use It
1. Board meeting prep. I upload the previous board report, current enrollment data, financial statements, leadership team meeting notes, and any relevant emails or memos. Then I ask NotebookLM to synthesize themes, flag items that need board attention, and draft talking points. What used to take three days takes about two hours. I wrote about this workflow in detail in my earlier post on NotebookLM and board reports.
2. Policy mastery. Upload your entire policy handbook — 200 pages, whatever. Then ask specific questions: "What does our policy say about staff use of personal devices for school communication?" or "Are there any conflicts between our data privacy policy and our social media policy?" It finds answers across documents that would take you an hour to locate manually.
3. Staff onboarding. Create a notebook with your faculty handbook, strategic plan, org chart, key procedures, and recent staff communications. New hires can ask questions and get source-grounded answers instead of waiting to find the right person to ask. With Google Classroom integration (launched September 2025), teachers can attach notebooks directly to Classroom assignments for students.
4. Research synthesis. Upload 10-15 research papers on a topic. Ask for a synthesis of findings, areas of agreement and disagreement, and gaps in the literature. The citations let you trace every claim back to a specific paper. I used this to build the evidence base for my AI Leadership course — 80+ sources synthesized into structured module content.
5. Course development. This is where I spend the most time. For a recent client project converting professional accounting lectures into AI avatar courseware, I uploaded chapter lectures, slides, and question banks, then used NotebookLM to identify topic coverage, map content to syllabus points, and spot gaps. For my own course, I use it to verify that every statistic in my scripts traces to a credible source. The citation system makes fact-checking fast instead of tedious.
Free Tier vs. Paid: What You Actually Need
| Feature | Free | AI Plus ($7.99/mo) | AI Pro ($19.99/mo) | AI Ultra ($249.99/mo) | |---|---|---|---|---| | Notebooks | 100 | 500 | 500 | 1,000+ | | Sources per notebook | 50 | 100 | 300 | 600 | | Daily queries | 50 | 100 | 500 | 5,000 | | Audio overviews/day | 3 | ~6 | 20 | 200 | | Words per source | 50,000 | 500,000 | 500,000 | 500,000 | | Video overviews | -- | -- | Standard | Cinematic |
My honest recommendation: Start with the free tier. 100 notebooks, 50 sources each, and 50 queries per day is enough for most people to get real value. The 50,000-word-per-source limit is the main constraint — if you're uploading long documents (100+ pages), you'll hit it.
If you use it regularly, AI Pro at $19.99/month is the sweet spot. 300 sources per notebook and 500 daily queries means you can build serious research environments. It also includes the upgraded NotebookLM Plus features. AI Ultra at $249.99/month is overkill unless you're running an entire department's research through it.
NotebookLM vs. Claude Projects: When to Use Which
I use both. They do different things well.
| | NotebookLM | Claude Projects | |---|---|---| | Source grounding | Strict — only your documents | Can synthesize beyond sources | | Hallucination risk | Near zero on your docs | Low but present | | Writing quality | Good | Best available | | Complex reasoning | Adequate | Superior | | Multimedia outputs | Audio, slides, mind maps, quizzes | Text, code, artifacts | | Free tier | Generous | Limited | | Ecosystem | Google Workspace | Standalone |
Use NotebookLM when you need verified, source-grounded answers. Policy questions. Fact-checking. Document comparison. Anything where accuracy matters more than creativity.
Use Claude when you need to write something, reason through a complex problem, or synthesize beyond what your documents explicitly say. Claude's writing quality and reasoning depth are still ahead.
Best practice: Use NotebookLM to verify what Claude tells you. Upload the same source documents to both, ask the same question, and compare. When Claude adds something that NotebookLM doesn't mention, that's either a useful synthesis or a hallucination — and now you know to check.
Limitations You Should Know
It only knows what you give it. This is the feature, but it's also the limitation. If your documents are incomplete or outdated, so are its answers. NotebookLM won't flag what's missing from your sources.
No academic citation formatting. It gives you inline numbered citations that link to source passages, but it won't generate APA, MLA, or Chicago-style bibliographies. If you need formal citations, you'll still have to create those yourself.
Audio overviews can oversimplify. The podcast format is engaging but it sometimes flattens nuance in complex material. Use Audio Overviews for orientation, not as your only engagement with the material.
Source limits exist. Even on the top-tier plan, you're capped at 600 sources per notebook and 500,000 words per source. For most people this is plenty. For truly massive document sets, you'll need to split across notebooks.
It's a Google product. Your documents are processed on Google's servers. Don't upload anything you wouldn't put in Google Drive. Apply the same data classification rules you'd use for any cloud tool — no student records, no confidential HR data, no information protected by legal privilege.
Start Here
Open notebooklm.google.com. Create one notebook. Upload the documents for whatever project is eating your time this week. Ask it three questions.
You'll know within ten minutes whether this tool is useful for your work. For most people who deal with documents daily, the answer is obvious.
I cover NotebookLM in depth in my AI Leadership course — specifically in the Context Engineering module (building AI environments with your own documents) and the Strategic Research module (using source-grounded tools to reduce hallucination in organizational workflows). If you want the full framework for integrating tools like this into your leadership practice, that's where to go.
