QUICK INFO
| Difficulty | Beginner |
| Time Required | 20-30 minutes for setup, ongoing for study sessions |
| Prerequisites | Google account, documents you want to learn from |
| Tools Needed | NotebookLM (free at notebooklm.google) |
What You'll Learn:
- Upload documents and structure them for effective learning
- Generate flashcards and quizzes grounded in your actual sources
- Use prompts that produce explanations at your level
- Build study sessions that adapt to what you don't know yet
This guide covers how to use NotebookLM's learning features with prompts that actually work. It's for anyone sitting on PDFs, lecture notes, or research papers who wants to study the material without rereading everything six times. If you just want to generate a podcast or summary, this isn't the guide for that.
Getting Started
Go to notebooklm.google and sign in with your Google account. The free version gives you 50 sources per notebook and 100 notebooks total, which is plenty for most learning projects.
Click "Create new notebook" and give it a name that reflects what you're studying. NotebookLM works best when notebooks are focused: one course, one research topic, one project. Dumping everything into a single notebook makes the AI's responses less targeted.
Adding Your Sources
Click "Add source" and upload your materials. NotebookLM accepts PDFs, Google Docs, Google Slides, text files, web URLs, YouTube videos with captions, audio files, Word documents, and Google Sheets. Each source can be up to 500,000 words or 200MB.
A few things the documentation doesn't make obvious: the tool only keeps a static copy of your files at upload time. If you update the original Google Doc, you'll need to manually re-sync it. Also, paywalled webpages don't work, and YouTube videos need captions (auto-generated is fine, but videos uploaded within the last 72 hours might not be available yet).
For learning purposes, I'd suggest uploading materials in logical chunks rather than one massive PDF. A chapter at a time lets you select specific sources when generating study materials, which keeps the output focused.
The Prompts
NotebookLM's chat and Studio features respond to your prompts, but they're grounded in the sources you've uploaded. That constraint is what makes it useful for actual learning rather than generic AI responses. The tool won't make things up that aren't in your documents, and it'll cite where each answer came from.
The Studio panel on the right side of the interface has built-in options for flashcards, quizzes, and reports. But the chat interface often gives you more control over the output.
Prompt 1: The Explanation Request
Explain [concept] as if I'm encountering it for the first time. Use examples from the source material, not generic ones.
This works because NotebookLM pulls from your uploaded documents specifically. If you're studying a technical manual, the examples will come from that manual. The "as if I'm encountering it for the first time" framing tends to produce clearer explanations than just asking "what is X."
When to use it: First pass through new material before diving into memorization.
Prompt 2: The Targeted Flashcard Generator
Create [15] flashcards focused on [specific topic or chapter]. Include the key terms I'd need to explain this topic to someone else.
The built-in Flashcards feature in Studio works fine for general decks, but specifying the topic and framing it around explanation rather than memorization produces more useful cards. You can also adjust difficulty by adding "focus on foundational concepts" or "include edge cases and exceptions."
Customize by: Changing the number and adding constraints like "no yes/no questions" or "each card should test understanding, not just recall."
Prompt 3: The Scenario Quiz
Create quiz questions that test whether I can apply [topic] to realistic situations, not just recall definitions. Present scenarios where I need to choose the right approach.
This prompt came from someone who was studying AI agent development and realized they could recognize terms but couldn't actually make decisions about when to use them. The scenario framing forces application-level questions rather than vocabulary tests.
Example output I got when testing this with a machine learning document: "A user uploads a 200-page PDF manual and wants to ask questions about it. Do you need: (a) fine-tuning, (b) RAG, (c) prompt engineering, or (d) function calling?"
Prompt 4: The Connection Finder
What concepts in these sources depend on each other? Show me which ideas I need to understand first before others will make sense.
This is useful when you're working through dense material and aren't sure what order to tackle things. The response usually maps out prerequisite relationships, which helps you avoid the frustration of reading about advanced topics before you've understood the basics.
Prompt 5: The Gap Identifier
Based on [topic], what questions should I be able to answer if I truly understand this material? List them from basic to advanced.
I find this more useful than just generating random quiz questions because it shows you the shape of what you're supposed to know. You can then self-assess against the list before generating actual quizzes.
Prompt 6: The Elaboration Request
Take [specific concept] and explain why it works the way it does, not just what it is. What problems would exist if it worked differently?
This prompt fights the tendency to learn definitions without understanding. The counterfactual framing ("what if it worked differently") pushes toward deeper comprehension.
Prompt 7: The Comparison Prompt
Compare and contrast [concept A] and [concept B] from the sources. Where do people typically confuse them?
Adding "where do people typically confuse them" surfaces the distinctions that actually matter for learning, not just surface-level differences.
Prompt 8: The Summary With Gaps
Summarize [section/chapter] but flag any points where the source material is ambiguous or where I might need additional resources to fully understand.
This is more honest than a standard summary. Instead of pretending everything is clear, it acknowledges where the source material has gaps. Useful for identifying what you need to look up elsewhere.
Prompt 9: The Teaching Simulation
I need to explain [topic] to [specific audience: a colleague, a student, my manager]. What are the three most important points to convey, and what's the simplest way to explain each?
The Feynman technique says you understand something when you can explain it simply. This prompt forces that exercise without requiring you to actually teach someone.
Using the Built-In Features
The Studio panel has dedicated buttons for Flashcards, Quizzes, and various report types. They're faster than writing prompts from scratch, and you can customize difficulty, topic focus, and card count before generating.
One feature worth knowing about: after a quiz session, you can generate follow-up materials based on questions you got wrong. That's the closest NotebookLM gets to spaced repetition, though it doesn't have true scheduling like Anki.
You can export flashcards as CSV files for import into dedicated spaced repetition apps if that's your workflow. The export option appears when you view a generated flashcard set.
The Learning Guide chat mode (accessible through chat settings) changes how NotebookLM responds. Instead of giving you answers directly, it asks probing questions and walks you through problems step-by-step. It's good for active learning sessions but slower than standard chat.
Troubleshooting
The flashcards are too basic: Add "focus on application and edge cases, not definitions" to your prompt. Or specify "difficulty: challenging" when using the built-in generator.
Responses cite sources I didn't want included: Before generating, select only the specific sources you want in the Sources panel. NotebookLM defaults to using all sources in a notebook.
The quiz explanations are too vague: Click "Explain" after answering a question. If that's still not enough, take the specific concept to the chat and ask for elaboration there.
Audio imports fail: Check that the file has clear speech. NotebookLM can't transcribe files with heavy background noise, overlapping voices, or no speech at all.
What's Next
Once you've got flashcards and quizzes working, look into the Audio Overview feature for passive review during commutes, or try the Video Overview for visual explanations of your sources. Both features are in the Studio panel and can be customized through the interface.
PRO TIPS
The keyboard shortcut Cmd/Ctrl+K opens quick search across all your sources. Use it when you half-remember where something was discussed.
When you're studying multiple related topics, create separate notebooks for each but use the same naming convention. NotebookLM doesn't have cross-notebook search, so organization matters.
If a source is too long for useful responses, split it into multiple smaller uploads. Chapters or sections work better than entire textbooks.
The chat history doesn't persist between sessions in the way you might expect. Save important generated content to your notes or export it before closing.
PROMPT TEMPLATES
Concept Deep Dive
I need to understand [CONCEPT] deeply enough to apply it, not just recognize it. Walk me through:
1. What it is in one sentence
2. Why it exists (what problem it solves)
3. How it relates to [RELATED CONCEPT]
4. A concrete example from the source material
Customize by: Replacing the concepts with whatever you're studying.
Example output: When I tested this with a prompt engineering document, it gave a structured breakdown that connected techniques to specific use cases from the uploaded material rather than generic explanations.
Exam Prep Generator
Generate a practice exam based on [SOURCE/CHAPTER] with:
- 5 recall questions (basic facts)
- 5 application questions (use the concept in a scenario)
- 2 synthesis questions (connect multiple concepts)
Provide an answer key at the end.
FAQ
Q: Can I use NotebookLM with sources in languages other than English? A: Yes. NotebookLM supports output in over 130 languages, and you can set the output language in the report settings. Audio imports work with many languages as well, though the full list varies.
Q: Does NotebookLM remember what I got wrong across sessions? A: Not automatically. You can generate follow-up materials based on missed questions within a session, but there's no built-in spaced repetition or progress tracking. Export to a dedicated SRS tool if you need that.
Q: What's the difference between the free and Plus versions for learning? A: Plus ($19.99/month with Google One AI Premium) increases sources per notebook from 50 to 300 and raises daily usage limits. For most individual learners, the free tier is sufficient.
Q: Can I share flashcard sets with classmates? A: Yes. Generated flashcards and quizzes can be shared via link. The person you share with doesn't need access to your original sources.
RESOURCES
- NotebookLM official site: Create notebooks and access your sources
- NotebookLM Help Center: Google's documentation on features and limits
- Supported source types: Full list of what you can upload
