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

NotebookLM complete tutorial, 2026 practical guide for organizing data and learning with AI

Every year when it’s time to prepare for exams or when it’s time to write a thesis, there’s always a pile of materials piled up on the desk. In the past, the only way to organize this information was to take notes manually, which was extremely inefficient. NotebookLM is an AI data organization tool developed by Google. It treats uploaded data as the only source of information, answers questions, generates summaries, and creates podcasts based on these data. This year, many students and professionals have begun to use it as a core tool for daily learning. This article will take you through this workflow from registration to advanced gameplay.

What exactly is NotebookLM?

NotebookLM was originally an experimental project of Google Labs and was later gradually opened to users around the world. Its core design concept is source-based answers, that is, it only uses the information you upload to respond to your questions, and will not randomly pull information online to fool you. This is completely different from asking questions directly with ChatGPT. The latter will make up freely, and every sentence given by NotebookLM will be marked with which paragraph of information is quoted.

Its model base is the Gemini series, which currently mainly supports the analysis of textual data, and also partially supports images in PDF. You can think of it as a small knowledge base that is exclusive to you, and all conversations will be based on the materials you give it.

How do domestic users access NotebookLM?

NotebookLM is currently only accessible through notebooklm.google.com, 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 problem yourself, otherwise the page will remain blank.

There will be a boot animation when you log in for the first time, just skip it. You will be asked to accept the privacy terms. By default, Google states that the uploaded data will not be used to train the model. This is important for people who deal with sensitive content. It is recommended to keep the default settings and do not actively check the data sharing options.

Create your first Notebook

Click New Notebook and the system will ask you to add sources first. The source type supports a wide range, including PDF documents, Google Docs, Google Slides, web links, copied and pasted plain text, YouTube video links, and audio files. A notebook can currently host dozens of sources. The capacity limit of the free version will be adjusted according to the policy. Please refer to the official public page for details.

After the upload is completed, each piece of information will be displayed in the left column. You can choose to let AI see only a few of them, or select all of them. In the middle is the conversation area, and on the right is the notes column, which can save useful answers in the conversation.

Data type supported range

Text-based PDFs are the best supported, and scanned PDFs can also be recognized. However, pure image formats sometimes lose words, so it is recommended to process them with OCR first. Google Docs links can be pulled directly and kept in sync, which means that after you modify the original document, the notebook will be updated accordingly.

Web links extract the entire page into text. If the original page has a lot of dynamically loaded content, it may be missing. In this case, it is recommended to print the page into PDF before uploading. YouTube videos will be transcribed into transcripts, and the accuracy depends on the clarity of the original video. Generally, English videos have the best effect, while Chinese video recognition will be worse.

Three core functions

The first is the abstract. Opening the notebook will automatically generate a summary of the material, including key topics, an outline of questions, and a list of suggested questions. This summary can be copied directly into Notion or Evernote as an overview of this batch of data.

The second is dialogue. You can ask any questions just like chatting with an AI assistant. The system will find relevant passage combinations from the data and mark the source number behind each fact point. Clicking on the number will jump to the original text, which is especially useful for academic writing because you can always check whether the citation is accurate.

The third is the audio overview, which is often referred to as the podcast function. The system will generate a conversation audio between two virtual hosts based on the data. The default is English, and currently supports multiple language switching including Chinese. One session lasts about ten minutes, which is suitable for listening to a brief summary of a long material while commuting.

Practical combat: Use NotebookLM to organize notes in a book

Take reading a 300-page English professional book as an example. The first step is to convert the e-book to PDF and cut it into several small files according to chapters and upload them separately, so that the scope of the dialogue can be selected more accurately. In the second step, let NotebookLM generate a summary for each chapter and go through it yourself to determine which are the core arguments. The third step is to ask questions about concepts you don’t quite understand and ask them to be explained with examples in the book.

The fourth step is to save the more exciting answers in the conversation as notes. These notes can be exported later. The fifth step is to let it generate a podcast version and listen to it repeatedly while walking or commuting to deepen your memory. After the entire process, a 300-page book can be digested in about a weekend, which is twice as efficient as manual note-taking.

Combine with other tools to build a complete workflow

NotebookLM itself is a relatively closed tool. Data can only be uploaded but not downloaded in batches. Conversations cannot export all records with one click. This is its biggest pain point. To solve this problem, you can use it in conjunction with other tools.

For example, organize and archive the daily conversations with AI using ChatGPT or Claude as a reference for future review. It is recommended to try here Save AI, this is a Chrome browser extension that can export conversations from more than a dozen AI sites such as ChatGPT, Claude, Gemini, etc. into PDF, Word, Markdown or long images. Local priority is given, and the data is not uploaded to the cloud. It is suitable to save today’s discussion with AI and directly drag it into NotebookLM tomorrow as a new source to continue questioning. This method of converting daily conversations into a long-term database is worth a try.

In addition, you can also use Obsidian or Notion as the main notebook, and NotebookLM as a special reading tool. Combining the two will bring out the best.

Prices, quotas and privacy

NotebookLM currently has a generous quota for the free version, which is more than enough for ordinary students and professionals for daily use. If you are a paid user of Google Workspace, you will automatically obtain NotebookLM Plus permissions, with higher capacity and function limits. Please see the official public page for specific quotas, because this part of the policy is constantly being adjusted.

In terms of privacy, Google clearly states that data uploaded by individual users will not be used to train models, but the policies for corporate accounts will be different. It is best to confirm with the IT department before handling sensitive materials within the company. Deleting a piece of data will remove it from your notebook, but the specific length of time that the data will be kept in the Google backend is not public. You should still be cautious when handling particularly confidential content.

Common misunderstandings

Many people who use NotebookLM for the first time expect it to be able to search the Internet like Perplexity, but they find that it will politely refuse to ask any questions beyond the information. This is not a bug, it is the original design intention. The purpose is to make the answers controllable and not divergent.

Another common misunderstanding is to upload dozens of materials and then ask them to answer everything at the same time. In this way, the system will be overwhelmed and the key points will not be grasped. It is recommended to only check the most relevant ones for each conversation, and the effect will be much better immediately.

Some people also complained that the tone generated by the Chinese podcast was a bit strange. There is indeed a gap between this part and the English version, and we look forward to subsequent optimization. But the actual information density is not bad, and you can get used to it after listening to it a few times.

FAQ

Is NotebookLM free?

Yes, a free version is currently available, which is basically enough for ordinary users for daily use. If you subscribe to Google Workspace or Google One AI Premium, you will automatically receive a higher quota for NotebookLM Plus. Please see the official public page for details.

Can NotebookLM be used in China?

直接打开 notebooklm.google.com 在国内是访问不了的,需要自己解决网络环境,并且要登录 Google 账号。Web 端体验最完整,移动端有官方 App 但功能略少。

上传的资料安全吗

Google 公开声明个人账户的资料不用于模型训练。但仍然建议处理高敏感材料前考虑脱敏,真正的核心机密文档不要上传到任何云端 AI 服务。

NotebookLM 和 ChatGPT 的区别在哪

NotebookLM 只用你给它的资料回答,会标注出处;ChatGPT 默认会自由生成,可能会编造。前者适合精准的资料整理,后者适合开放的创意和讨论。两者搭配使用效果最好。

播客功能可以用中文吗

可以,目前已经支持中文音频概览,但音色和语气和英文版相比还在持续优化中。中文版适合作为听觉辅助,精确逐字稿建议参考原资料。