NotebookLM is built on Google's Gemini large language model and is deeply integrated with Google Drive and Google Workspace. It accepts sources including Google Docs, Google Slides, PDFs, web URLs, YouTube video links, and plain text, up to 50 sources per notebook with up to 500,000 words each. The system uses retrieval-augmented generation (RAG) to ensure all generated content is grounded in the provided context rather than the model's general training data. Its Audio Overview feature uses text-to-speech synthesis to create two-host conversational summaries of uploaded materials, and a NotebookLM Plus (paid) tier offers higher usage limits and customization options.