NotebookLM complete usage tutorial, 2026 practical guide for using AI to organize data and study

A Complete NotebookLM Tutorial: The 2026 Hands-On Guide to Organizing Materials and Studying with AI

Every year when exam season or the thesis deadline comes around, a stack of materials piles up on the desk. In the past, organizing these materials could only be done by hand, and the efficiency was extremely low. NotebookLM is an AI material-organizing tool from Google; it treats the materials you upload as the sole information source, answering questions, generating summaries, and producing podcasts around them. This year many students and professionals have begun adopting it as a core tool for daily study; this article takes you through the entire workflow from registration to advanced play.

What Exactly Is NotebookLM?

NotebookLM started as an experimental project of Google Labs and was gradually opened up to users worldwide. Its core design philosophy is source-grounded answering—that is, it uses only the materials you upload to respond to your questions, and it will not casually pull information off the internet to fob you off. This is completely different from asking ChatGPT directly, which freely makes things up, whereas every sentence NotebookLM gives cites which passage of material it came from.

Its model foundation is the Gemini series; it currently mainly supports parsing text-type materials and also partly supports images within PDFs. You can think of it as a small knowledge base exclusively yours, where every conversation unfolds based on the materials you give it.

How Users in China Can Access NotebookLM

NotebookLM can currently only be accessed at notebooklm.google.com; it requires a Google account and must be used in an environment where the network can access Google services. If you are in China, you need to solve the network-environment issue yourself, or the page will keep showing a blank screen.

After logging in, the first entry shows a guided animation, which you can skip. The system will ask you to accept the privacy terms; by default Google states that uploaded materials will not be used to train the model, which is important for people handling sensitive content. We recommend keeping the default settings and not actively checking the data-sharing option.

Creating Your First Notebook

Click to create a new notebook, and the system will have you add sources first. The range of supported source types is fairly broad, including PDF documents, Google Docs, Google Slides, web links, copy-pasted plain text, YouTube video links, and audio files. A single notebook can currently hold dozens of sources, and the free version's capacity limit will be adjusted with policy—refer to the official public page for specifics.

After uploading, the left sidebar displays each piece of material, and you can choose to let the AI see only a few of them, or select all. The center is the conversation area, and the right side is the notes panel, where you can save the more useful answers from a conversation.

The Range of Supported Material Types

Text-type PDFs are the best supported; scanned PDFs can also be recognized, but pure-image formats sometimes lose characters, so it is best to run them through OCR first. Google Docs links can be pulled directly and kept synced, which means after you modify the original document, the notebook updates along with it.

A web link extracts the entire page into text for use; if the original page has a lot of dynamically loaded content, it may be missing, in which case we recommend printing the page to PDF first and then uploading. A YouTube video gets transcribed into a text script, with accuracy depending on the clarity of the original video; English videos usually give the best results, and Chinese video recognition is worse.

The Three Core Features

The first is summarization. Opening a notebook automatically generates a summary of the materials, including key themes, a question outline, and a list of suggested questions. This summary can be copied directly into Notion or Yinxiang Note (Evernote China) as an overview of this batch of materials.

The second is conversation. You can ask follow-up questions just like chatting with an AI assistant, and the system finds relevant passages from the materials to compose an answer, marking a source number after each fact point. Clicking a number jumps to the original text, which is especially useful for academic writing because you can verify the accuracy of a citation at any time.

The third is the audio overview, also commonly known as the podcast feature. The system generates an audio conversation between two virtual hosts based on the materials; it defaults to English and currently already supports switching among multiple languages including Chinese. A segment is about a dozen-plus minutes, suited to getting the gist of a long piece of material on your commute.

Hands-On: Using NotebookLM to Organize Notes for a Book

Take reading a 300-page English professional book as an example. Step one, convert the e-book to PDF and cut it into several small files by chapter to upload separately, which lets you more precisely select the conversation scope. Step two, have NotebookLM generate a summary for each chapter, then go through it yourself to judge which are the core arguments. Step three, ask follow-up questions about concepts you do not quite understand, asking it to explain by combining examples from the book.

Step four, save the more brilliant answers from the conversation as notes, which you can later export in bulk. Step five, have it generate a podcast version and listen to it repeatedly while walking or commuting to deepen your memory. With the whole flow done, a 300-page book can basically be digested in about a weekend, with efficiency double that of pure manual note-taking.

Building a Complete Workflow with Other Tools

NotebookLM itself is a relatively closed tool: materials can only be uploaded, not batch-downloaded, and conversations cannot be exported in full with one click, which is its biggest pain point. To solve this, you can string it together with other tools.

For example, organize and archive the conversations you have with AI via ChatGPT or Claude each day, as a reference for later lookups. Here we recommend trying Save AI, a Chrome browser extension that can export conversations from a dozen-plus AI sites such as ChatGPT, Claude, and Gemini into PDF, Word, Markdown, or long images. It is local-first, with data not going to the cloud, suited to saving today's discussion with AI and dragging it directly into NotebookLM tomorrow as a new source to keep asking follow-ups. This way of distilling everyday conversations into a long-term material library is worth a try.

Beyond that, you can use Obsidian or Notion as your main notebook and NotebookLM as a dedicated reading tool, combining the two to each play to their strengths.

Price, Quota, and Privacy

NotebookLM currently has a quite generous quota even on the free version, more than enough for ordinary students and professionals in daily use. If you are a paid Google Workspace user, you automatically get NotebookLM Plus access, with higher capacity and feature limits; refer to the official public page for the specific quota, because this part of the policy keeps being adjusted.

On privacy, Google clearly states that materials uploaded by individual users will not be used to train the model, but the policy for enterprise accounts differs, so it is best to confirm with the IT department before handling sensitive internal company material. Deleting a piece of material removes it from your notebook, but how long it is retained on Google's backend is not publicly disclosed, so still be cautious when handling especially confidential content.

Common Misconceptions

Many people, the first time using NotebookLM, expect it to be able to search the web like Perplexity, only to find it politely refuses any question outside the materials. This is not a bug; it is the design intent, aimed at keeping answers controllable and non-divergent.

Another common misconception is uploading dozens of materials and then asking it to answer everything together, which drowns the system and instead causes it to miss the key points. We recommend checking only the few most relevant materials per conversation, and the results will improve immediately.

Some people also grumble that the tone of the Chinese podcast generation is a bit odd; there is indeed still a gap from the English version here, and we look forward to future optimization. But the actual information density is not bad, and after a few listens you get used to it.

Frequently Asked Questions

Is NotebookLM free?

Yes, it currently offers a free version, basically enough for ordinary users' daily use. If you subscribe to Google Workspace or Google One AI Premium, you automatically get the higher quota of NotebookLM Plus; refer to the official public page for specifics.

Can NotebookLM be used in China?

Opening notebooklm.google.com directly is not accessible in China; you need to solve the network-environment issue yourself and log into a Google account. The web experience is the most complete; there is an official mobile app, but with slightly fewer features.

Are uploaded materials safe?

Google publicly states that materials from individual accounts are not used for model training. But we still recommend considering redaction before handling highly sensitive material, and truly core confidential documents should not be uploaded to any cloud AI service.

What is the difference between NotebookLM and ChatGPT?

NotebookLM answers using only the materials you give it and cites sources; ChatGPT by default generates freely and may make things up. The former suits precise material organization, the latter suits open creativity and discussion. Using the two together works best.

Can the podcast feature use Chinese?

Yes, it currently already supports a Chinese audio overview, but the voice and tone are still being continuously optimized compared with the English version. The Chinese version suits being an auditory aid; for a precise word-for-word transcript, we recommend referring to the original material.