The Chat section¶

The central column, called Chat, is the space for interacting with artificial intelligence. Here you ask questions, request analyses, and receive responses based exclusively on the sources selected at that moment in the Sources column, formatted according to your request. All interactions (user questions and system responses) are automatically saved and remain available in subsequent sessions unless explicitly deleted by the user.

As already mentioned, the system only allows interaction with sources uploaded by the user, which provides very strong control over so-called hallucinations. In fact, when asked questions about topics not contained in the sources, NotebookLM responds that it cannot fulfill the request due to lack of information. This approach reduces the risk of unverifiable information, though it should be remembered that verification of AI-produced content is always the user's responsibility.

The choice of sources is entirely the user's responsibility: their validity determines the quality of responses.

Interface structure¶

Starting from the top, in the upper right band you can see two icons: - Three vertical dots: allows complete deletion of the entire chat, except for the summary that's automatically created when sources are first loaded. It's not possible to delete a single response. This can be useful for managing space in a frequently queried notebook, where the risk is ending up with a very long chat - perhaps save the most useful responses in the Studio area first. - Three horizontal lines: opens a popup to configure the chat

This popup allows you to configure the entire chat, acting on two aspects:

• Behavior: you can work on objectives, style, and roles that NotebookLM should maintain in responses. The available choices are:

  • Default: for general research and brainstorming activities.
  • Learning guide: for educational content, allows understanding new concepts and skills effectively through an interactive process with the user.
  • Custom: allows you to easily define a true system prompt for the notebook, up to 10,000 characters. These instructions can include:
    • The level of language formality
    • Preferred response structure (e.g., bullet points vs. narrative paragraphs)
    • Use of specific terminology or industry jargon
    • Analysis approach (e.g., favoring synthesis or detailed exploration)
    • Emotional tone (e.g., neutral, enthusiastic, cautious)
    • Specific perspectives to adopt or emphasize

  Thanks to this advanced customization, you can create tailored conversational experiences for different purposes, from analyzing scientific literature to explaining complex concepts for beginners, from exploring ideas to critically evaluating arguments.

• Response length: you can choose from:

  • Shorter: Concise responses that get straight to the point, ideal for quick reviews or when only essential information is needed
  • Default: A balance between detail and conciseness, suitable for most situations
  • Longer: Detailed and extensive responses that thoroughly explore topics, ideal for complete analyses or in-depth learning

This setting influences all responses generated in the Chat column, allowing you to adapt system verbosity to the specific context you're operating in, whether it's a quick information check or an immersion in a complex topic.

In the central part of the column, user questions are visible aligned to the right, and system responses aligned to the left. At the bottom of the column is the space to enter the prompt.

When sources are first loaded, a general notebook summary is generated, with a concise synthesis of the main topics covered in the loaded sources. If additional sources are loaded later, this summary is not updated.

This summary cannot be removed; the only way to eliminate it is to delete all sources.

Below the summary, as below every NotebookLM response, you'll find some buttons for specific functions:

• Save to note: saves the element to the Studio section
• Copy icon: copies the element to clipboard
• Thumbs up and down: allow you to rate the text quality

The main part is dedicated to the actual conversation: you type questions in the input area at the bottom and view responses in the upper part. Each response includes numerical references (like [1], [2]) that link statements to original sources. Clicking on these numbers takes you to the source reference for the information used at that point.