browser · idea · product

NotebookLM (Google)

Source-grounded study tool that turns your documents into audio and video overviews, mind maps, quizzes and reports.

inspireevidence: mixedby Google Labsnotebooklm.google ↗

NotebookLM (Google)

  • Maker: Google (Google Labs)
  • URL: https://notebooklm.google
  • Status (10 October 2026): shipping, free with paid tiers. Audio Overviews (2024) made it famous; mind maps arrived in 2025; Video Overviews launched 29 July 2025 (narrated slides with images, diagrams, quotes and numbers pulled from the sources), with a Studio panel holding audio, video, mind map and report outputs, several per notebook. One secondary source reports “Cinematic Video Overviews” in March 2026 (Gemini 3, Veo 3, Ultra plan only); we couldn’t confirm it from Google.

What it is

A research and study tool grounded in sources you upload: it answers questions with citations and turns the sources into other formats (podcast-style audio, narrated slide videos, mind maps, study guides, flashcards, quizzes).

The problem it’s solving

Reading and digesting a pile of documents is slow, and one format doesn’t suit every learner or every moment (commuting, reviewing, first pass).

Its path / bet

Grounding first (answers come from your sources, with citations), then many generated representations of the same material. The visual side is a set of fixed output types the model fills, not free-form UI.

How it works (concretely)

Gemini models over the uploaded sources. A mind map is a navigable tree of topics from the selected sources; a Video Overview is a generated slide deck with narration, steerable by focus, learning objective, audience and expertise.

Strengths

  • Grounding and citations make it trustworthy in a way free generation isn’t.
  • Fixed output types (map, deck, audio) are reliable and familiar.
  • Very large adoption; the Audio Overview format was widely copied.

Weaknesses / limits

  • Mostly passive formats (listen, watch). The mind map is the only manipulable one, and it’s navigation, not a model you can change.
  • No evidence we found that the formats improve learning, as opposed to convenience.

Relation to fictty

Inspire. Two lessons: ground the representation in real sources, and offer a few fixed, well-made formats rather than infinite generated ones. Both match fictty’s catalogue-plus-data-sources design. Different medium and market.

Could fictty adopt it instead of building?

No. A closed Google product (third-party MCP wrappers exist, unofficial).

What fictty should take from it

  • Every number on a teaching screen should trace to a source. NotebookLM’s citations are the same principle as fictty’s data sources; show the source of a bound value on demand (the command it came from).
  • The mind map is a good component to have: a navigable tree with a detail pane. fictty has list, table and text; a tree view is worth checking for in the component set.

Sources