How to use NotebookLM in 2026 (the beginner's guide nobody made simple)

Picture this: you've got a 40-page research paper, three hours of meeting transcripts, and a YouTube lecture you never had time to watch. Normally that's a weekend of reading and note-taking. With NotebookLM, you can dump all of it into one place and have an AI that's read every word of it, ready to answer questions, summarize, quiz you, or even turn the whole thing into a podcast you can listen to on a walk.

That's the pitch. And unlike a lot of AI tools that promise the moon, NotebookLM mostly delivers — because it's built around one simple rule: it only talks about what you feed it.

What exactly is NotebookLM?

NotebookLM is Google's source-grounded AI notebook, now folded into the broader Google AI ecosystem (you may also see it referred to as "Gemini Notebook" in some Google documentation). You upload documents, videos, or audio files as "sources," and the AI reads everything and grounds its answers in those documents only, it doesn't make things up from the open web, as one detailed 2026 guide to the tool puts it.

That grounding is the whole point. Every answer comes with inline citations pointing back to the exact passage it pulled from — you can click a citation and jump straight to that spot in your source. It's less "chat with an AI" and more "interrogate your own research library with superhuman recall."

Google's official Help Center is the best place to bookmark for how-tos, and the Learn about Gemini Notebook page covers the core feature set in more depth than most third-party tutorials.

How do you upload sources into NotebookLM?

Head to notebooklm.google.com, sign in with a Google account, and click the "+" button to start a new notebook. From there, you can add sources several ways:

1. Upload files directly — PDFs, Google Docs, Word docs (.docx), plain text, Markdown, CSV, and PowerPoint files are all supported
2. Paste a website URL — NotebookLM scrapes the text content of the page
3. Paste a YouTube link — it pulls in the video's transcript, not the audio or visuals
4. Upload audio files — MP3 and WAV formats work as sources
5. Paste copied text directly into a source box

A few quirks worth knowing before you start uploading everything on your desktop. Only public YouTube videos with captions, either user-uploaded or auto-generated, are supported, and videos without speech aren't supported at all. Paywalled webpages won't work either, and if a source page uses only images with no text layer (like a scanned PDF), NotebookLM won't be able to read much of it.

Google's guide to adding sources has the full list of supported file types and edge cases, which is worth a skim if you're planning to feed it anything unusual.

What are the free tier limits you need to know?

This is the part most tutorials get wrong or leave outdated, so here's what's actually current as of 2026. Each individual source can contain up to 500,000 words or up to 200MB for uploaded files, and that ceiling stays the same no matter what plan you're on — upgrading raises how many sources you can add, not how big a single one can be.

For the free tier specifically, you can include up to 50 sources per notebook, with 100 notebooks allowed per account. Daily usage caps also apply — the free plan generally allows around 50 chats, 3 Audio Overviews, and 3 Video Overviews per day, though Google adjusts these numbers periodically, so it's worth checking current quotas directly on the Gemini Notebook help page if you're running a heavy workload.

If you outgrow the free tier, NotebookLM Plus now comes bundled with Google AI Plus, which raises sources to 100 per notebook and lifts the daily chat and generation limits. Higher tiers (Pro and Ultra) push those numbers even further, up to 600 sources per notebook on Ultra — but for a normal study load, a single class, or a focused research project, you're unlikely to hit the free ceiling.

How do you generate an Audio Overview?

This is NotebookLM's signature feature, and it's genuinely impressive the first time you hear it. Once you've added sources, look for the Studio panel and select Audio Overview. NotebookLM generates a conversational, podcast-style discussion between two AI hosts who walk through your material like they're genuinely curious about it — not reading a summary out loud.

You can choose different formats depending on what you need. Use Deep Dive for connected explanation, Brief for a sub-two-minute summary, Critique for constructive evaluation, and Debate for multiple perspectives, according to one detailed breakdown of the current format options. You can also add a custom prompt to steer the tone — ask for something aimed at "a general audience" versus "engineers," for instance, and the hosts will adjust accordingly.

Can you actually talk back to the podcast?

Yes, and it's the feature that separates NotebookLM from a plain summarizer. Interactive Mode lets you join newly generated Audio Overviews and ask questions with your voice — you interrupt the hosts mid-conversation, ask your question, and they answer using your sources before picking the conversation back up. One important catch: interactive mode only works on newly generated overviews in the Deep Dive format, so if the Join button is missing, you need to generate a fresh Deep Dive overview rather than reusing an old one.

It genuinely changes how you consume audio content — instead of rewinding to catch something you missed, you just ask.

What other outputs can you create from your sources?

The Studio panel isn't just audio. From the same set of sources, NotebookLM can generate study guides, briefing documents, FAQs, timelines, flashcards, quizzes, and mind maps that map out how concepts in your sources connect to each other. There's also a Video Overview option that turns your material into a narrated slideshow with diagrams and visuals, plus newer additions like Slide Decks and Infographics for turning dense material into something presentable.

Because every output is generated from the sources you selected, you can toggle sources on or off using checkboxes in the source list to control exactly what feeds into a given output — handy when you want a quiz based only on chapter 3, not your entire textbook.

What's the difference between NotebookLM and ChatGPT or Claude?

This is the question beginners ask most, and it comes down to one design decision. ChatGPT and Claude answer from broad general knowledge by default, pulling from everything they were trained on unless you explicitly paste in context. NotebookLM flips that: you give it a source first, then it helps you ask questions, synthesize information, and turn that material into actual usable outputs — it won't wander off into general knowledge or guess at things your sources don't cover.

That trade-off matters depending on what you're doing. Need creative brainstorming, code help, or general advice? A general chatbot is more flexible. Need to know exactly what your contract says, what a specific paper found, or what was discussed in six months of meeting notes — with a citation proving it — NotebookLM is built for that job specifically.

How should students, researchers, and professionals each use it?

Students get the most obvious win: upload a textbook chapter, lecture slides, and a recorded class, then generate a study guide and a quiz before an exam. Ask for a Brief audio overview on the commute, then a Deep Dive with Interactive Mode when you actually sit down to study.

Researchers benefit from dropping in a stack of papers at once and asking NotebookLM to identify contradictions, common themes, or gaps across all of them simultaneously — something that would take hours of manual cross-referencing otherwise.

Professionals tend to get the most value from less glamorous sources: policy documents, meeting transcripts, onboarding materials. Upload a faculty handbook or a year of board meeting minutes, and instead of digging through folders, you just ask when a specific policy was last discussed.

What are the best quick tips to get more out of it?

A couple of small habits make a real difference once you're past the basics. First, convert any web search result NotebookLM surfaces directly into a source, so it's part of what future answers draw from. Second, get comfortable toggling sources on and off — narrowing which documents the AI is reading often produces sharper, more specific answers than throwing everything at it every time.

NotebookLM isn't trying to replace ChatGPT or Claude, and it shouldn't. It's solving a narrower, more useful problem: giving you an assistant that actually read the thing you gave it, and can prove it. Start with one messy pile of PDFs sitting on your desktop right now — you'll have a study guide, a summary, and a podcast about it before you'd have finished skimming the first document manually.