How to Learn Anything With AI

The bottleneck in learning something new was never access to the material. It was that the material only came in one format, and the format rarely matched the moment. NotebookLM changes that: you upload what you want to learn, and it reshapes the same source into a podcast, a mind map, flashcards, a quiz, an infographic, or a slide deck. Below is the full workflow - three steps, eight output modes, and the order to run them in. We build learning systems like this for 300,000+ readers at AI Central, and the rest of the collection is in the AI Central Library.

Everything downstream depends on what you put in. NotebookLM works from sources you supply rather than general knowledge, which is what makes its output trustworthy enough to study from.

A source can be a YouTube video, a book, a transcript, an article, a whitepaper, a PDF, or pasted text. A single notebook holds up to 300 sources depending on your plan, so you are not limited to one document per topic - you can build a notebook that covers an entire subject.

Upload it, and the notebook is ready.

If your source has gaps, NotebookLM can fill them. It searches the web for missing context and adds new sources to the notebook alongside what you uploaded.

Two settings. Use Fast Research for quick inputs when you need a bit more context around a specific point. Use Deep Research when you want a full report and a broader set of sources pulled in.

This step is optional for a reason. If you uploaded a book, you probably do not need it. If you uploaded a single article, you almost certainly do.

This is where the tool earns its place. The same source can come back in eight different shapes, and the skill is matching the mode to the moment rather than defaulting to whichever one you tried first.

A structured explanation of your source, delivered as a narrated video. Best for focused sessions where you are sitting down with the intention of learning something properly.

Your source turned into a conversational podcast. This is the mode that converts dead time into study time: walking, commuting, training, cooking. Nothing else on the list works while your hands are busy.

The structure of the material, laid out visually. Best used first rather than last - seeing how the pieces connect before you go deep tells you what you are actually looking at, and takes seconds.

A study guide built from your source: short-answer questions, essay prompts, and a glossary of key terms. This is the mode that turns passive material into something you can be tested on.

Active recall, generated automatically. Rereading a chapter feels productive and does very little; being asked a question and failing to answer it is what moves material into memory. Flashcards are the cheapest way to run that loop.

Multiple-choice testing with immediate feedback. The value is diagnostic: it finds the gaps in your understanding while they are still small enough to close in an afternoon.

The key ideas in a single scannable visual. Not a learning tool so much as a refresher - the thing you look at before a meeting, six weeks after you did the real work.

One slide per concept. The underrated use here is teaching: presenting material is the hardest test of whether you understand it, and a generated deck gives you the scaffold to try.

Run in isolation these modes are novelties. Run in sequence they behave like a course:

One thing worth naming, because it decides whether any of this works. Modes 1, 2, 3, and 7 are input - you receive material. Modes 4, 5, 6, and 8 are output - you produce something and get corrected. Nearly everyone runs the input modes, enjoys them, and stops. The learning happens in the second group.

And if you are still getting oriented with AI tools generally, start with our beginner guide to generative AI.

NotebookLM is Google's research and study tool. You upload your own sources - documents, videos, articles, transcripts - and it generates summaries, audio and video overviews, mind maps, flashcards, quizzes, infographics, and slide decks grounded in that material rather than in general knowledge.

PDFs, documents, text files, audio files, images, Google Docs and Slides, website links, YouTube videos, and pasted text. A notebook holds up to 300 sources depending on your plan, so a full subject can live in one place.

There is a free tier that covers the core features with lower source and generation limits. Paid Google plans raise those limits, which matters mainly if you are building large notebooks or generating output heavily.

Flashcards and the quiz, by a distance. Audio and video overviews are pleasant and get you oriented, but recall testing is the part that moves material into memory. Use the overviews for intake and the testing modes for actual learning.

For getting oriented quickly, largely yes. For material you need to genuinely know, no - it changes the format, not the effort. The workflow above is designed around that limit, which is why half the modes make you produce something. More AI learning systems are in the AI Central Library.