Most interactions with large language models feel like a conversation with a knowledgeable generalist. You ask a question, and the model draws upon its vast, pre-trained knowledge to provide an answer. NotebookLM operates on a fundamentally different principle: source grounding. Instead of searching the entirety of its internal knowledge, it focuses exclusively on the documents you provide.

This deliberate constraint is the architecture's core strength. It transforms the AI from an open-ended conversationalist into a focused research assistant. By limiting the model's context to your specific materials, you create a closed system for reasoning. This drastically reduces the risk of hallucinations because the model isn't allowed to invent facts or pull from unrelated data; it must base its answers on the evidence you've supplied. Every response is tied directly back to your source material, often with citations.

Each notebook you create is a completely separate workspace, like a dedicated project folder. The sources uploaded to a notebook about quarterly financial reports will never be referenced when you're working in another notebook about historical biographies. This structure allows you to build distinct, highly-specialised knowledge hubs without any risk of cross-contamination.

Think of it this way: a standard chatbot is like a librarian in a massive, public library. They can answer questions on almost anything, but their knowledge is broad. A NotebookLM notebook is like hiring a private researcher, giving them a specific stack of books and reports, and telling them, "For this project, this is the only information that exists."

This architectural choice ensures that the AI's expertise is tailored precisely to your needs for a given task. You are not just prompting a model; you are curating its world.

The NotebookLM interface is designed around this source-grounded architecture. It consists of two primary areas that work in tandem.

1. The Source Sidebar: This is your notebook's library. On the left side of the screen, you'll see a list of every document, link, or text you've added. This isn't just a file list; it's the complete set of facts the AI is allowed to use.
2. The Interaction Area: This is the main workspace where you chat with the model and take notes. Your queries here prompt the model to synthesise information from the documents in the source sidebar.

Critically, you can interact at two different levels. By default, your questions are posed at the notebook level, meaning the AI considers all uploaded sources to formulate its response. However, you can also select a single source. When a source is selected, your questions become source-level, and the AI will only use that specific document to answer. This allows for both broad, cross-document analysis and deep, focused inquiry into a single text.

Now, let's test your understanding of these core architectural concepts.

Understanding this structure is the key to using NotebookLM effectively. You are not just a user of the AI; you are the architect of its knowledge.