NotebookLM's Latest Features Are Insane

NotebookLM is positioning itself as a full research workflow for academia—turning a curated set of sources into chat answers, audio/video explainers, mind maps, study tools, and even presentation-ready visuals, all with citations back to the underlying documents. The core value is that questions aren’t answered in isolation; they’re grounded in the notebook’s uploaded and discovered sources, then returned with traceable references so researchers can verify what came from where.

The process starts with building a notebook and adding sources. Users can upload up to 50 sources for free (and up to 300 with a paid plan), then organize them into smaller “projects” such as chapters or paper sections. A standout addition is source discovery: NotebookLM can search for additional materials that may have been missed, effectively acting like a mini literature review to sanity-check whether the source list is complete. After sources are assembled, NotebookLM generates a shared context—what the materials have in common—then supports a chat experience that draws on the full set of sources (the transcript describes a workflow using organic photovoltaics as an example).

In chat mode, the output is presented as a referenced response, with the ability to click through to see which file(s) supplied specific information. The chat also includes controls for conversational goals (such as learning-guide style), response length, and custom instructions—so the same source base can be used for different research tasks, from quick orientation to deeper synthesis.

Where the tool goes beyond typical “AI Q&A” is in its knowledge-to-study transformations. NotebookLM can save chat outputs into notes for later retrieval. It also expands the earlier audio feature into configurable audio overviews and AI podcasts, with options like deep dives, brief critiques, debates, language selection, and host focus areas. Interactive audio mode lets users join in with follow-up questions.

Video overviews add a visual layer, offering explainer-style or bite-size summaries with multiple visual styles. The transcript highlights that these outputs can be scrubbed through like a structured presentation, and that the generated visuals are detailed enough to feel “pleasant to watch,” not just functional.

For studying and entering a field, NotebookLM generates mind maps that break down topics into granular subareas—useful for spotting what a newcomer should know before discussions with experts. It can also produce quizzes and flashcards based on the notebook’s sources, including question prompts about specific technical details (e.g., disadvantages of certain materials as transport layers).

Finally, NotebookLM’s beta infographic and slide-deck features aim at dissemination: users can generate editable infographics, slide decks, and longer reports such as study guides or technical reviews. The transcript emphasizes that these outputs include diagrams and structured text that would otherwise take hours to assemble manually.

Data privacy is addressed with claims that NotebookLM doesn’t use content for generative AI model training unless feedback is shared, and that uploads, queries, and responses aren’t visible to Google—an important consideration for academic workflows. The transcript closes with practical recommendations: upload papers to draft sections of a peer-reviewed manuscript, and—if allowed—upload MP3 recordings of meetings to generate summaries, next steps, and email-ready recaps grounded in the discussion.