notebooklm

---
name: notebooklm
description: Use when the user wants anything done in NotebookLM — reading a notebook, adding a source, generating Studio outputs, or creating a notebook; guidance only, no browser control.
category: Research
version: 1.0.0
tools: []
---

# NotebookLM Guidance

Guide a user through getting the most out of Google NotebookLM (notebooklm.google.com) — Amara has no browser control, so this is advisory: ask what the user wants, then hand them exact steps and, for Studio outputs, a ready-to-paste custom prompt. Never claim to have performed an action inside NotebookLM.

## Intake (up to 4 questions, one at a time, most stop at 3)

1. **What do you want to do?** Read/extract (ask a question of an existing notebook), add a source (URL, text, file, Google Doc, or synthesized content), generate a Studio output (Audio/Video Overview, Mind Map, Report, Flashcards, Quiz, Infographic, Slides — the exact set varies by account and rollout), or create a new notebook. Refuse to give generic advice ("just open NotebookLM") — pin down the action first.
2. **Which notebook** (for read/add-source/generate)? A name or a URL — if it's ambiguous, tell the user to check the homepage grid for the right card.
3. **Action-specific detail:**
   - *Read/extract:* the exact question to type into the notebook's chat.
   - *Add source:* which type — URL/website/YouTube link, copied text, file upload, Google Doc, or synthesized content (pre-processed elsewhere then pasted as "copied text").
   - *Generate:* which Studio output, plus the angle/audience/length — **default Studio prompts produce mediocre output; a custom prompt is what makes it good.**
   - *Create new:* the title, and any initial sources to add.
4. **(Studio outputs only) Full custom-prompt detail** — angle, audience, length. This becomes the exact text the user pastes into the customization menu next to the Studio button (not the main generate button).

## Advice by Action

**Read/Extract:** Tell the user to open the notebook, type the question into chat using natural phrasing (NotebookLM's chat handles it best), and wait a few seconds for the response.

**Add sources:** Match the source type to the right sub-flow — Add Source → Link for URLs/YouTube, Add Source → Copied Text for pasted text, the file picker for uploads, Add Source → Google Docs for Drive files. For a **synthesized-content add** (the most powerful pattern): rather than feeding NotebookLM a raw noisy URL, pre-process the content yourself first — pull the main article, strip navigation/ads/comments — then have the user add it as "Copied text." This produces dramatically better summarization than raw ingestion. After any add, tell the user to wait for the ingestion spinner to clear before continuing.

**Studio outputs:** NotebookLM's Studio panel inventory changes with account tier and rollout — as of the last check (2026-06) it included Audio Overview, Video Overview, Mind Map, Reports (Briefing Doc, Study Guide, FAQ, Timeline, custom formats), Flashcards, Quiz, Infographic, and Slides, but treat that list as a hint, not ground truth; tell the user to check the live panel (including any "discover more" overflow) for what's actually available to them. The critical instruction: **open the customization menu next to the output button — not the main button — and paste in a detailed custom prompt** built from the angle/audience/length gathered at intake. Examples:

- **Audio Overview:** "Two-host conversation between a researcher and an experienced practitioner. Audience: non-technical executive making a budget decision. Length: 8-10 minutes. Focus on business implications, not technical depth. Include one concrete example per major point. Acknowledge counter-arguments briefly."
- **Infographic:** "Decision-tree style. Action-oriented (each panel ends with a decision or action). 6 panels max. Monochrome navy + amber highlight. Each panel: title (4-6 words), 1-2 sentence body, decision/action line. No filler panels."
- **Study Guide:** "Undergraduate-level (define every technical term). Structure: 6 concepts × 4 elements each — definition / why it matters / one worked example / 3 practice questions. Practice questions should be apply/analyze level, not recall."
- **Slides:** "12 slides max. 1-2 sentences per slide body. Presenter notes per slide with one concrete example, one likely audience objection, and how to address it. No bullet points in slide bodies — prose only. End with a one-slide call-to-action."

**Create new notebook:** Title first, then add initial sources via the same sub-flows as above; the auto-summary that follows is usually fast (well under a minute).

## Timing — Set Expectations Correctly

Studio generations vary enormously in wall-clock time and NotebookLM notifies in-app when ready, so tell the user not to sit and wait for the slow ones:

| Output | Typical time |
|---|---|
| Add a source | Seconds to ~30s |
| Chat response | A few seconds |
| Audio Overview | 5-10 minutes |
| Video Overview | 5-15 minutes |
| Infographic / Slides / Mind Map | 2-5 minutes |
| Study Guide / Briefing Doc / FAQ / Flashcards / Quiz | ~30-60s |
| New-notebook auto-summary | Under 30s typically |

For anything in the multi-minute range, tell the user to start it and check back — NotebookLM will notify them — rather than waiting live.

## Practical Notes

- Login walls: if the user isn't signed in, that's on them to resolve — never suggest sharing credentials.
- If a notebook name is ambiguous, tell the user to check the homepage grid rather than guessing which one is meant.
- If a Studio button isn't visible, tell the user to check for a "discover more" overflow before concluding the feature isn't available to their account.
- If the user wants the output saved somewhere, remind them NotebookLM hosts its own outputs (in-app playback, downloadable infographics, etc.) — Amara can't fetch or store the resulting file, only relay the report back in chat.

## Why the Questions Matter

Each action takes a different path through NotebookLM's UI and needs different parameters, so pinning the action down first (question 1) avoids sending the user down the wrong flow. The notebook identity (question 2) matters because ambiguous names cost the user a disambiguation round-trip that a URL avoids entirely. The action-specific detail (question 3) exists because each source type and each Studio output goes through its own sub-flow — for Studio outputs specifically, the default prompt behind the main button is consistently mediocre, so skipping the customization step is the single most common way users get a disappointing result. The full custom-prompt detail (question 4) is what actually becomes the text pasted into that customization box, so vague answers here ("make it good") produce vague generations.

## Error Handling

| Situation | Response |
|---|---|
| User just says "open NotebookLM" | Refuse to guess; ask what action they want. |
| Ambiguous notebook name | Ask them to confirm via the homepage or provide the URL. |
| Studio output type not visible in their panel | Suggest checking overflow/"discover more"; note the feature may not be enabled for their account/tier yet. |
| User wants Amara to click through the UI | Clarify Amara has no browser control — offer the exact steps and prompt text instead. |