Source: raw/Gemini_Notebook_NotebookLM_is_becoming_the_Most_Essential_AI_Tool_7_Use_Cases_to_Try.md (https://www.youtube.com/watch?v=8nzt_xzsyq8)
Creator: Grace Leung
Platform: YouTube
Duration: 18:15
Published: 2026-01-16
Skill level: Beginner to intermediate (no code; web app only)
Grace Leung walks through seven marketing uses of Google NotebookLM, which she calls her favourite AI tool because it answers only from your sources (“minimal hallucination”) and can now turn those sources into slide decks, infographics, videos, data tables, quizzes and podcast-style audio. The recurring trick is a loop: generate something, save it as a note, convert the note into a source, then select only that source for the next output. The last use case hands a whole notebook to Gemini to build an interactive page. The video was recorded in January 2026. The current title says “Gemini Notebook”, which matches the July 2026 rename reported in Claude Code Driving Creative Tools, but the transcript only ever says NotebookLM. ^[inferred: the title was probably updated after upload] The video is sponsored by HubSpot (a free “AI Toolkit I use every week” guide).
Key Takeaways
- Three panels: Sources (she says at least 50 sources can be uploaded), Chat (prompt against your sources) and Studio (one-click outputs).
- Built-in research when you have no sources. Fast research is a quick scan; deep research runs a full analysis, writes a report and finds many more sources. Tell it your research focus, which reputable sources to prefer (she named Forrester and Gartner, and it used them) and, most importantly, the timeframe. Import the found sources in one click.
- Ask for the questions, not a summary. “What are the five most important questions a marketing leader should be asking about content strategy this year?” surfaces priorities that challenge your thinking. A gap-analysis prompt (“what sources are missing that could be critical?”) critiques the source set itself. Save good answers as notes even though chat history exists.
- Keep reusable specs as notes. Write a “slide deck specification” note (audience, slide count, structure, style), then paste it into the Slide Deck prompt along with the content focus. She recommends the presenter slides format. The decks follow the spec’s colour scheme and structure; she credits the Nano Banana image model for the visuals.
- Custom persona per notebook. Configure the notebook with a role (her example: a growth strategist with project context, tone and priorities). Answers then follow the persona, flag risks and ask clarifying questions.
- Save, convert to source, regenerate. Report → export to Google Docs → reimport the Doc as a source → select only it → generate a process infographic in the same brand style. The same loop turns drafted infographic copy into a source for a cleaner infographic.
- Brand characters work, with caveats. Upload a mascot image as a source and tell the infographic prompt to use it throughout. NotebookLM has a tight safety filter around characters (faces, copyrighted designs), so results are hit or miss. For consistent character use she switches to the Gemini app or Google AI Studio, “same model, but it follows the character details more reliably.”
- Data tables turn messy sources into a spreadsheet. Describe the columns, generate, and export to Google Sheets in one click.
- NotebookLM holds the facts; Gemini builds. In the Gemini web app, Add file → NotebookLM attaches a whole notebook. With Canvas on, Gemini built her a working landing page from a notebook in about five minutes. Her framing: NotebookLM is “your second brain,” Gemini is “your builder.”
The Seven Use Cases
- Research synthesis into a leadership deck. Example: content-marketing trends for annual planning with 30+ industry reports and no time to read them. Deep research → strategic and gap-analysis questions → slide-deck spec note → presenter slides.
- Client-project knowledge base. Drop in call transcripts, scope documents, strategy decks, performance data, research and handwritten-note images (Google Drive integration keeps them organised). Set the persona, then use Report → Project overview (she advises leaving its default prompt alone) and export to Google Docs. Add a launch-roadmap infographic in the client’s brand colours, and an SOP built with the create your own report option. Share the notebook so teammates can catch up.
- Branded social infographics. Bulk-import YouTube links as sources (paste a list of URLs), upload the brand character, then generate infographics. She specifies type, hand-drawn style and “use this character throughout.” For a second version: draft room-by-room copy in chat, save it as a note, convert it to a source, then generate in square orientation with a whiteboard-marker style and grid layout.
- Learning a field fast. Use Perplexity to find current sources (URLs only), bulk-import them, generate a mind map first for the landscape, then a customised audio overview. You can set the format (e.g. deep dive, critique), length, audience, focus proportions and how the hosts interact. Interactive mode lets you ask the hosts questions, answered from the sources. The mobile app plays the same notebook’s audio on the go.
- Training package from one source set. Her example imports 50 official Claude Code documentation pages. Ask chat for a two-hour workshop outline (she got six parts with learning objectives) and save it as a note. Then produce a video overview (explainer format, custom visual style, outline pasted into the prompt), a presenter slide deck with a narrator character (a friendly bear; her attempt at the Claude mascot came out as a cloud character), and a quiz with set question count, difficulty and focus topic.
- Competitive messaging research. Ask Google AI Mode for the top 20 email-marketing platforms as URLs only (her logic: these are the brands AI search already recommends, so their messaging is worth studying). Bulk-import them, generate a data table (tool name, headline, target audience, value proposition), export to Sheets, then reimport the Sheet and generate a competitor-landscape infographic grouped by audience.
- Notebook to interactive app via Gemini. Attach the notebook in Gemini, turn on Canvas and prompt for a fully functional landing page with a stated visual style and contents. Her sleep-science demo produced animated stars, a hero logo, sleep-cycle explainer, myth flip-cards, a self-assessment tool and a chatbot widget. She also suggests dashboards and apps as outputs.
Try It
- Spec notes first. Before generating any deck, create a note with audience, slide count, structure and style, and reuse it every time.
- Run the competitor table in 20 minutes. AI Mode list of top competitor URLs → bulk import → data table with your own columns → export to Sheets. This doubles as a check on which brands AI search recommends in your category.
- Use the save-as-source loop whenever an output needs to follow specific copy: draft in chat → save as note → convert to source → select only that source → generate.
- If a mascot gets filtered, move the same prompt to the Gemini app or AI Studio rather than retrying in NotebookLM.
- Pair with AEO work. Grace’s own AEO playbook uses NotebookLM to analyse the pages an AI answer cites and draft a better-sourced piece.
Implementation
Tool/Service: NotebookLM (renamed Gemini Notebook, July 2026, per the wiki’s existing coverage); Gemini web app (Canvas mode); Google Docs/Sheets/Drive; Perplexity and Google AI Mode for source finding.
Setup: New notebook → add sources (upload, Drive, pasted URL list, or Fast/Deep research) → optionally configure a custom persona → use Chat for analysis and Studio for outputs (Slide Deck, Infographic, Report, Data Table, Mind Map, Audio Overview, Video Overview, Quiz).
Cost: Not stated in the video.
Integration notes: Reports export to Google Docs and data tables to Google Sheets; both can be reimported as sources. Infographics take language and orientation settings plus a style prompt. Gemini reads a whole notebook through Add file → NotebookLM. For driving NotebookLM from Claude Code instead of the web UI, see Claude Code Driving Creative Tools (the notebooklm-py route).
Open Questions
- Source limit. She says “at least 50 sources” per notebook; the actual cap by plan is not stated and is not verified here.
- Caption errors left in place: “task book illustration style” (probably “textbook”, unconfirmed); “Germinal app” is the Gemini app (certain from context).
- Timing. These are January 2026 UI paths. The June 2026 video (Google AI content engine) shows notebooks accessible inside the Gemini app, and the product has since been renamed, so menu labels may have changed.
- “Minimal hallucination” is her characterisation of source-grounded answers, not a measured error rate.
Related
- Claude Code Driving Creative Tools — running NotebookLM decks, podcasts and mind maps from Claude Code; source of the Gemini Notebook rename date
- Google AI Content Engine (Grace Leung) — her June 2026 follow-up using notebooks inside Gemini as a brand knowledge base
- Nano Banana — the image model behind NotebookLM’s slide decks and infographics
- AEO Playbook (Grace Leung) — NotebookLM as a citation reverse-engineering tool
- Cowork + Skills Content System (Grace Leung) — the same creator’s Claude-side content system for comparison