How to Use NotebookLM (Google's AI "Tool for Understanding")
NotebookLM is Google's source-grounded AI tool that lets you upload up to 20 documents and have a conversational AI draw exclusively from those sources. It collapses reading, research, note-taking, and writing into a single integrated workspace, eliminating the need for manual organization. ---
Key Concepts
| Concept | Definition |
|---|---|
| Source-grounded AI | The AI only draws from documents you explicitly provide — no hallucinations from the open web, no mixing with other users' data |
| Notebook | A self-contained workspace for a single project; holds all sources, conversations, and pinned responses |
| Context window | NotebookLM supports up to 20 sources × 200,000 words each = 4 million words total — larger than any other AI tool at time of recording |
| Pin board | A persistent space at the top of the interface where you save ("pin") AI responses you want to keep visible |
| Citations | Every AI response links back to the exact passage in the source document, shown highlighted in context |
| Tools for thought | The broader category of software (including second brains) aimed at augmenting human cognition — the conceptual lineage NotebookLM belongs to |
Notes
Getting Started
- Visit `notebooklm.google.com`; sign in with a Gmail account
- At time of filming: limited to US users aged 18+
- Create a new notebook with the `+` button and give it a project-specific title
- Three main interface areas:
- **Left column** — Sources panel
- **Bottom box** — Chat/conversation input
- **Top area** — Pin board for saved responses
Adding Sources
- Three source types: Google Drive documents, PDF uploads, copy-pasted text
- Select up to 20 documents; each can be up to 200,000 words
- Clicking a source shows an AI-generated summary and key topics, plus full scrollable content
- You can add a new inline note at any time via "Add note" — paste text from anywhere (email, web, another app)
Core Workflow
- Load multiple related sources to enable cross-document synthesis
- Ask questions conversationally; follow up to expand, distill, reformat, or redirect
- Use suggested questions (pre-generated, context-specific) as jumping-off points
- Pin responses you want to keep; otherwise the conversation is ephemeral
- Citations let you trace every claim back to its source passage with surrounding context
Use Case: Understanding Difficult Material
- Upload a complex PDF (e.g., academic paper)
- Ask plain-language questions to get summaries and explanations
- Drill into specific confusing points with follow-up questions
- Ask it to reformat findings into a letter, memo, or argument for a specific audience
Use Case: Meeting Notes / Transcripts
- Upload a full meeting transcript (tested with a 26-page / 12,000-word Google Doc)
- Ask for main points, decisions, or action items
- Extend further: ask it to draft a proposal, curriculum, or follow-up document based on the conversation
- Safe to upload proprietary content — sources are not used to train any model and are wiped when you leave the notebook
Use Case: Writing — Editorial Mode
- Upload a draft as a source
- Ask for improvement suggestions, then ask for a full rewrite following those suggestions
- Add a second source (e.g., notes from a book on writing craft) and ask it to rewrite the draft applying advice from that book
- Citations show exactly which advice from the reference book was applied
Use Case: Writing — Ideation Mode
- Start with one source (e.g., book notes) and ask for related ideas or research avenues
- Add a second source and ask for connections between the two books
- Narrow toward a specific angle, add more sources, ask for article title ideas → outline → opening lines
- Demonstrates going from no idea → compelling opening paragraph through iterative conversation
Getting Your Notes into NotebookLM via Readwise
- **Readwise** (paid) syncs highlights from Kindle, Readwise Reader, and other sources to Google Docs
- Go to Readwise → Export → connect Google Docs (same account as NotebookLM)
- Key setting: keep "send each source to its own Google Doc" **enabled** so you can selectively choose individual books/articles as sources
- If disabled, all highlights merge into one or a few docs — useful for conversing with your entire reading history at once but harder to filter
- Run the export; a `readwise` folder appears in Google Drive, ready to use as NotebookLM sources
Limitations
- Experimental platform — bugs and learning curve expected
- Cannot connect directly to note-taking apps (Evernote, Notion, Google Keep) — must use an intermediary like Readwise
- Hard limit of 20 sources per notebook
- Struggles with math, messy PDF formatting, and precise detail retrieval in very large sources
- Team confirmed fixes and improvements are in progress
Actionable Takeaways
- Create a NotebookLM account at `notebooklm.google.com` and start one notebook for a current project
- Connect Readwise to Google Docs to automatically export book/article highlights as individual Google Docs, making them ready for NotebookLM
- When uploading a long transcript or document, ask for main points first, then progressively ask for deliverables (proposal, curriculum, letter)
- Use the **pin board** immediately — save any response worth keeping before continuing the conversation
- Add a writing-craft reference (e.g., notes from a book on writing) as a second source when revising drafts to get style-informed rewrites with citations
- Keep the "one source per Google Doc" setting enabled in Readwise to stay within the 20-source limit while retaining selectivity
Quotes Worth Keeping
You point the AI at a specific set of documents, effectively giving it instant expertise in whatever knowledge they contain — so you always know it's only drawing on authoritative, private, trusted information.
Until now, the stages of the creative process — reading, researching, note-taking, and writing — were completely separate. NotebookLM gets all those steps and folds them into a single integrated space so you can remain in that wonderful flow state and keep moving forward.
It's an AI collaborator trained on the data only you possess, with your unique view of the world.