NotebookLM is an AI-powered research and note-taking tool from Google, built on the Gemini model family, that lets a user upload their own sources, such as PDFs, web pages, and Google Docs, and then ask questions, request summaries, or generate study guides drawn only from that uploaded material rather than the model's general training knowledge. It grew out of an earlier internal project called Project Tailwind, first previewed in 2023, and was publicly rebranded and expanded as NotebookLM in 2024.
History
Project Tailwind began as an experimental effort within Google Labs to address a common weakness of general-purpose chatbots: giving confident but ungrounded answers, or Hallucination (AI), when asked about specific documents. By restricting the model's responses to a user-supplied set of sources and citing passages within those sources, the tool aimed to make its outputs more verifiable. Google renamed the product NotebookLM in mid-2023 and continued a limited rollout before a wider public release in 2024, when it also introduced the "Audio Overview" feature: an automatically generated, podcast-style conversation between two AI voices summarizing and discussing the uploaded material. The Audio Overview feature drew significant public attention and viral social-media interest for its naturalistic tone, including moments where the AI hosts appeared to reflect on their own artificial nature, and became NotebookLM's most widely discussed capability.
Features
NotebookLM's core mechanism is a form of Retrieval-augmented generation: rather than relying purely on parametric knowledge learned during training, it retrieves relevant passages from the user's uploaded sources and constructs answers with citations back to those passages, a technique intended to improve Grounding (AI) and reduce fabricated claims. Beyond question answering, it can generate study guides, FAQs, timelines, and briefing documents from a notebook's sources, and later versions added support for a wider range of file types and larger numbers of sources per notebook, drawing on Gemini's growing Context window capacity. The Text-to-speech pipeline behind Audio Overviews synthesizes two distinct AI voices engaging in an unscripted-sounding discussion, a capability built on Google's broader work in Multimodal AI generation.
Reception
NotebookLM was generally well received by researchers, students, and knowledge workers for reducing the tedium of synthesizing long documents, and the Audio Overview feature in particular was credited with introducing many casual users to practical Large language model applications beyond chatbots. It also drew scrutiny typical of grounded AI tools: while source-restriction reduces certain categories of hallucination, it does not eliminate the risk of the model misreading or overgeneralizing from the provided material, and some users noted the audio hosts could still make confident-sounding errors about their source documents. Google continued to expand NotebookLM's availability and features through 2024 and 2025, integrating it more closely with Google Workspace and positioning it as an example of applied Semantic search and document understanding built on its Gemini models.