Today, at White Rabbit Ashland, I gave a talk (and demo) about how I use Gemini Notebook (formerly known as Google NoteBookLM). Below you will find the notes from my presentation. Note: Come back periodically to this blog post as I update the [TODO] sections over the next few days after tonights talk.
Top 11 reasons why I use Gemini Notebook (instead of chatbots)
- I can curate the sources.
- Works for a wide range of use cases.
- I can audit/validate the sources.
- Supports the sources types I want to use / YouTube, research articles, tech news.
- Has Python programming in a secured sandbox – [TODO: add section showing how to upload a CSV file and request graphs and analysis of the data. Follow on TODO – create custom Python code to process the CSV file data]
- Generates the types of digital assets I can use for real work – blog posts, infographics, videos
- Great at summarizing content – pope’s AI 94 page document to 4 pages
- Infographics are great visual summaries to add to blog posts and articles
- Video overviews for YouTube video posts (now with shorts)
- Discord server for feedback and feature requests.
- Google Pro level gives me what I need
Use Cases for Google’s Gemini Notebook
Use Case 1: Strategic “Red-Teaming” and Finding Hidden Blind Spots
Most organizations use AI defensively to summarize what is already written inside their strategy documents. A far more powerful workflow involves turning Gemini Notebook into an offensive “gap finder” to identify what is critically missing .Rather than feeding the platform complementary or agreeing files, this counter-intuitive workflow relies on uploading conflicting or competing source files —such as a legacy regulatory inspection report alongside a newly drafted organizational improvement plan. By forcing the AI to bridge the distance between where an enterprise was instructed to improve and where its current strategy claims to go, leadership teams can uncover dangerous operational misalignments in seconds.
The 4-Phase Audit Protocol
To execute a strategic gap audit, supply the notebook with two opposing document sets and run a structured, four-phase prompt sequence:
- Requirement Match: Prompt the AI to cross-reference every explicit mandate or historical area for improvement from the legacy audit report against the new strategic plan, verifying whether every required item is explicitly addressed.
- Evidence Interrogator: Prompt the AI to evaluate internal impact measures, separating concrete quantitative metrics from purely descriptive statements.
- Cohort Stress Test: Instruct the AI to isolate specific sub-groups, demographic targets, or localized compliance metrics to expose zero-mention omissions.
- Devil’s Advocate: Direct the AI to adopt the persona of a rigorous external auditor, generating high-challenge interview questions complete with inline citations pointing to discrepancies between past findings and current goals.
Real-World Audit in Action
Consider a practical example evaluating an educational institution (“St. Algorithm’s International”) comparing its legacy inspection report against its new School Improvement Plan (SIP):
- Cohort Stress Test Discovery: Running Phase 3 flags that while past inspections highlighted Key Stage 2 boys’ writing stamina as a critical deficiency, the new SIP mentions writing stamina metrics zero times , focusing exclusively on reading comprehension.
- Evidence Interrogator Discovery: Running Phase 2 exposes that a high-profile £15,000 library investment is backed only by descriptive statements regarding pupil premium progress, lacking verifiable quantitative metrics to prove direct educational impact to inspectors.By utilizing this four-phase protocol, organizations can “red-team” their own operational strategies on a Tuesday afternoon, identifying critical blind spots internally rather than explaining strategic omissions to external auditors or clients on a Wednesday morning.”Illuminate then act. Use the AI to shine a light on the gap, but use your professional judgment to fill it.”
Use Case 2: Autonomous Goal-Driven Research and Interactive Reports
Gemini Notebook allows users to initiate comprehensive research projects even when starting with zero uploaded documents. Rather than requiring users to manually assemble a folder of PDFs before asking questions, the platform incorporates agentic web-gathering tools: Fast Research and Deep Research .
- Fast Research: Users input a broad query or target topic (e.g., “Docs about Q4 planning” or “Most Important Fossil Finds” ). Gemini Notebook queries Google Drive or the public web, presenting a curated list of relevant source candidates complete with titles, links, and brief descriptions explaining how each item aligns with the prompt. Users can manually review and select specific items to import into their grounded context.
- Deep Research: For complex topics, this agentic feature autonomously browses up to hundreds of websites, analyzes disparate findings, and generates detailed multi-page research reports in minutes. Alongside the report, it presents a complete index of relevant sources—both cited and uncited—allowing users to individually review, deselect, or import materials to build a permanent source library.
From Static Summaries to Interactive Reports
Once sources are gathered, Gemini Notebook can synthesize these materials into structured Interactive Reports . Rather than delivering a flat summary, an Interactive Report organizes source material into a navigable learning experience featuring a left-hand thematic navigation sidebar, embedded mind maps, and downloadable slide deck structures.Crucially, Interactive Reports ground abstract concepts in concrete operational reality. For example, an Interactive Report generated across 22 technical papers on industrial AI agents avoids generic definitions like “An AI agent is an autonomous system.” Instead, it leads with a practical industrial scenario: an AI model predicting a boiler feed pump alert that was muted by an operator, ultimately causing catastrophic pump failure. This shifts the platform from a system where users must bring their own material into a goal-driven partner that builds and organizes a verifiable knowledge base from scratch.
Use Case 3: Grounded Writing Directly Inside Google Docs
For freelancers, SEO agencies, and enterprise teams, switching back and forth between research archives and writing applications introduces heavy cognitive friction. Google Workspace integration bridges this gap by embedding Gemini Notebook directly into Google Docs.By opening the Gemini side panel or bottom bar within Google Docs, writers can tag an entire curated notebook using the @ symbol (e.g., @Market_Research_Notebook). Once tagged, the document editor gains direct access to that notebook’s grounded source library.
Eliminating the “Platform Tax” in Client Workflows
- Tab-Free Drafting: Writers can prompt Gemini to generate document sections, executive summaries, or case studies directly inside Docs, using the tagged notebook as the explicit knowledge base.
- Automated Inline Citations: Statements generated in Docs automatically include clickable inline citations referencing the original source document stored in the notebook.
- Crucial Configuration Preference: To guarantee factual traceability, writers must verify that automatic inline citations are toggled ON under preferences within the Gemini Notebook interface. If toggled off, drafts return clean but unsourced; when enabled, every claim links directly to a verified passage.This integration delivers massive commercial leverage for agencies and solo operators. An SEO consultant working with a local service business (e.g., a dental practice or roofing contractor) can assemble a single 20-source notebook containing local competitor pricing, keyword research, service pages, and customer testimonials.That single reusable research asset can subsequently power ten local blog posts, case study introductions, and email sequences directly inside Docs without paying recurring “platform taxes” for separate research tools or subscription content software.
Use Case 4: Automated Data Analysis via Sandboxed Python Cloud Execution
Standard AI language models struggle with mathematical calculations and dense tabular data because they operate on statistical text token prediction rather than precise computation. Gemini Notebook resolves this limitation through an integrated, secure cloud computer attached to each notebook sandbox.When a user prompts Gemini Notebook to analyze numerical datasets—such as 100,000-row CSV files, Census Bureau state population tables, or financial tables buried inside scanned PDFs—the platform writes and executes native Python code in an isolated cloud environment behind the scenes.
Quantitative Computation vs. Qualitative Text Analysis
- Calculations & Data Cleaning: Instead of estimating numerical outcomes, the cloud environment executes precise Python math to compute averages, calculate variable correlations, identify data anomalies, and aggregate subset totals.
- Structured File Outputs: The Python engine executes data transformations and delivers downloadable output files directly into the Studio Panel. Users can export generated PNG trend charts, SVG vector diagrams, formatted Excel (.xlsx) spreadsheets, PDFs, JSON files, and Word documents.
- Synthesizing Text & Numbers: A single notebook can host dense qualitative literature alongside quantitative CSV files. Users can execute queries that synthesize textual research findings with computed mathematical trends simultaneously.
Use Case 5: Hands-Free Voice Coaching, Lecture Recording, and Mobile Workflows
The Gemini Notebook mobile application (available on iOS and Android) extends grounded research into hands-free and on-the-go environments through two primary mobile features:
1. Real-Time Spoken Voice Conversations
Users (aged 18+) can engage in two-way, spoken audio dialogue with their notebooks across nearly 100 languages. Because every spoken response is grounded entirely in the uploaded sources, users can utilize the mobile app as an interactive voice tutor or professional briefing assistant.Unlike static audio playback, users can interrupt the AI spoken output mid-sentence—asking for clarification, requesting a simpler explanation, or directing the AI to focus strictly on a specific sub-folder of sources (e.g., “Wait, stop—explain that concept again using only my onboarding notes” ).
2. Mobile Audio Recorder
The mobile app features a built-in live audio recorder. When attending a live lecture, client meeting, or recording walk-and-talk thoughts during a commute, the audio stream is automatically transcribed into text upon saving. The resulting transcript is immediately appended to the notebook as an editable text source, complete with citation support for future querying and drafting.
Use Case 6: Blending Published Literature with Personal Notes via “Expert Intelligence”
A central challenge in research is combining personal notes with copyrighted, authoritative literature without violating intellectual property boundaries or relying on unverified web scrapes. Google’s Expert Intelligence initiative solves this by integrating Google Play Books directly into Gemini Notebook.Users can import eligible purchased Google Play Books—drawing from a catalog of over 100,000 titles from major publishers—directly into their source panel alongside personal class notes, meeting transcripts, or internal standard operating procedures (SOPs).
Architectural and Legal Framework
- Query-Time Retrieval Layer: Rather than relying on static model training memory, Gemini Notebook retrieves the full licensed text of the purchased ebook at the exact moment a query is submitted. This creates a legal, publisher-backed licensed content layer operating via query-time retrieval rather than model pre-training.
- Copyright Enforcement in Shared Workspaces: To protect intellectual property rights, if a user shares a notebook containing a Google Play Book source with a colleague or student, the recipient must also individually own an eligible copy of the ebook to view, query, or interact with that specific source text.This capability allows students to combine leading textbooks with class syllabi for exam preparation, or corporate managers to synthesize published management best practices alongside internal performance data.
Use Case 7: Instant Multimedia Content Engine: Podcasting, Slides, and Short Videos
The Studio Panel acts as a multi-format publishing engine, transforming raw source text into auditory, visual, and computational artifacts.
| Studio Asset | Key Technical Features & Styles | Practical Business & Academic Use Case |
| Audio Overviews | Two AI hosts engage in a natural podcast-style discussion. Supports four formats: Deep Dive , Brief , Critique , and Debate . Includes interactive interruption for live Q&A. | Hands-free learning during commutes, screen-free research reviews, and audio study guides. |
| Short Video Overviews | Generates ~60-second vertical visual explainers combining narrative AI voiceovers with educational animations and slide graphics in 80+ languages. | Rapid team onboarding clips, social media authority posts, and quick executive summaries. |
| Slide Decks | Automatically builds complete presenter presentations from grounded sources. Supports post-generation slide revisions via targeted prompts. | Preparing client decks, lecture outlines, and team meeting presentations from raw research. |
| Infographics | Single-page visual summaries offered in distinct visual styles ( Instructional , Whiteboard , Watercolor , Retro Print , Kawaii , Anime ). | High-level topic overviews, cheat sheets, visual documentation, and social media graphics. |
| Data Tables | Extracts all tabular data across selected sources into a single structured table. Features an “Export to Sheets” option routing data and citations to Google Sheets. | Literature reviews, competitive matrix analysis, price comparisons, and data normalization. |
Use Case 8: Personal Interactive Study Suite and Team Onboarding System
Gemini Notebook converts passive reading materials into an active, self-testing learning environment powered by Google’s Learn LM model family.
Active Recall & Knowledge Transfer
- Interactive Mind Maps: The system constructs visual node-based maps of interconnected source concepts. Clicking on any individual node allows users to query the notebook specifically about that single thematic sub-branch without sifting through the broader document pool.
- Adaptive Quizzes & Flashcards: Users can generate flashcards or quizzes targeting specific weak areas or entire notebooks. Supported quiz formats include Multiple Choice , Multiple Select , Short Answer , and Fill-in-the-Blank . Questions can be manually edited or expanded post-generation.
- Post-Quiz AI Coaching: Following the completion of a quiz, users can chat directly with the notebook regarding their test results. The AI analyzes missed answers and provides targeted, source-cited coaching to address specific knowledge gaps.
- Enterprise SOP Onboarding: Organizations can upload internal handbooks, standard operating procedures, and product documentation to generate interactive training suites for new hires—replacing flat PDFs with interactive testing modules and voice-driven Q&A.
Feature Requests
You can use the Gemini Notebook Discord server to give feedback, bug reports and feature requests.
- Videos using my voice
- Multi-presenter video overviews (audio overviews have 2 person conversation).
- Programming API programming languages.
- Language Translations for reports.
- Cross notebook joins (you can add sources to multiple notebooks).
- Mobile app – delete sources UI needs improvements.
What file formats can be exported from Gemini Notebook?
Gemini Notebook allows you to export deliverables across several formats depending on whether you are generating assets through Studio, exporting directly to Google Workspace, or running code in the Python sandbox:
- Cloud Code Execution & File Deliverables
When Gemini Notebook runs Python code in its sandboxed environment to analyze data or build custom outputs, you can generate and download the following file formats:
Visual Charts & Graphics: PNG (.png) and SVG (.svg).
Spreadsheets & Datasets: Excel (.xlsx) workbooks and CSV (.csv) data files.
Documents & Written Reports: PDF (.pdf), Microsoft Word (.docx), Markdown (.md), and JSON (.json) files.
Presentations: Microsoft PowerPoint (.pptx) decks.
- Direct Google Workspace Exports
For text notes, reports, and extracted data tables, the three-dots menu provides one-click exports:
Google Docs: Export written reports, notes, and study summaries directly into a new Google Doc.
Google Sheets: Export structured data tables directly into a new Google Sheet.
- Studio Multimedia Formats
Audio Overviews (Podcasts): Downloadable as MP3 audio files to listen to on the go.
Video Overviews & Shorts: Downloadable MP4 video files featuring narrated animations.
Infographics, Mind Maps & Slide Decks: Visual outputs and slide decks generated in the Studio panel can be exported or downloaded as image/presentation assets.
What source document types can Gemini Notebook process?
Gemini Notebook supports a broad array of document types, multimedia files, and web sources. Every source added is grounded with traceable inline citations back to the original text or media passage. The following is the list of supported source document types by category:
Documents & Office Files
PDF Files (.pdf)
Microsoft Word (.docx)
Google Docs (auto-updated and synced from Google Drive)
Plain Text (.txt) & Markdown (.md) files
Spreadsheets & Tabular Data
CSV Spreadsheets (.csv)
Google Sheets (up to 100,000 tokens per file)
Presentations
Microsoft PowerPoint (.pptx)
Google Slides (up to 100 slides)
Books & E-books
ePub Files (.epub)
Google Play Books: Eligible purchased e-books imported directly from your Google Play library
Audio Recordings
Audio files are automatically transcribed upon import into searchable text sources.Supported Audio Extensions: MP3, WAV, M4A, AAC, OGG, FLAC, AIFF, WMA, OPUS, AMR, 3GP, AVI, MPEG, and more.
Images (supported Formats)
Images provide visual context for queries and analysis: Supported Extensions: AVIF, BMP, GIF, HEIC, HEIF, ICO, JP2, JPE, JPEG, JPG, PNG, TIF, TIFF, and WEBP.
Web, Video & Interactive Content
Web URLs: Scraped text content from public HTML web pages.
YouTube URLs: Text transcripts extracted from public YouTube videos with captions.
Copied & Pasted Text: Custom text snippets pasted directly into the upload box.
Gemini App Chats: Conversation logs saved directly into your notebook from the Gemini app.
What are the source limits?
Gemini Notebook enforces the following source and notebook limits:
Per-Source Limits
- Word Limit: Up to 500,000 words per individual source.
- File Size: Up to 200 MB for local file uploads.
- Google Slides: Up to 100 slides per presentation.
- Google Sheets: Up to 100,000 tokens per spreadsheet.
Notebook Source Capacity
- Free Plan: Up to 50 sources per notebook.
- Paid Subscriptions: Up to 300 sources per notebook.
Gemini Research Modes
Fast Research and Deep Research serve distinct research workflows in Gemini Notebook, depending on whether you need quick file lookups or an exhaustive automated synthesis.
Comparison Summary
| Feature | Fast Research | Deep Research |
| Speed & Processing | Delivers instant search results. | Operates in the background for a few minutes while analyzing sources. |
| Search Scope | Searches either the Web or your Google Drive. | Autonomously browses and examines up to hundreds of websites across the web. |
| Primary Use Case | Quick lookups, finding specific Drive files, or adding targeted web links. | Starting research on a brand-new topic from scratch or conducting in-depth market/academic surveys. |
| Output Generated | A list of matching search results with titles and brief relevance descriptions. | A structured multi-page research report plus a complete list of all underlying sources. |
Key Differences
Autonomy & Depth: Fast Research acts as a rapid lookup tool that presents matching pages or files for your review. Deep Research functions as an agentic assistant that reads through dozens or hundreds of web pages, evaluates the findings, and synthesizes them on its own.
Google Drive Integration: Fast Research allows you to search your Google Drive using natural language queries (such as “slides from Claire”) alongside web search. Deep Research focuses primarily on deep web-wide exploration.
Report Generation: Fast Research returns standalone source links for manual selection. Deep Research automatically compiles a multi-page synthesis report that you can import into your notebook alongside the individual source URLs.
Cautionary Note: you should always check the returned links before you choose to add the URLs to your sources.
Power Tip: How to quickly add a YouTube playlist of videos into one of your notebooks.
I wanted to quickly add all of the YouTube Stanford University lectures done by computer scientist Donald Knuth. I found the following playlist. The trick was to find a way to quickly get all of the lecture URLs from the playlist. By searching on Google I found the following Vercel app URL “How to extract URLs from any YouTube or YouTube Music playlist”:
https://export-youtube-playlist.vercel.app/tools/youtube-playlist-link-extractor
Use the following playlist URL in the playlist extractor and click the “Extract Links” button : https://www.youtube.com/playlist?list=PL94E35692EB9D36F3
After the extraction completes (some confetti will fly to alert you) and get the following result (with the URLs to more than 100 of his YouTube lecture videos):
Click the “Copy All” button. Note: you can also use the “Export As” button to create CSV,TXT and other text files containing the list of video URLs.
You can then quickly add all of his lectures to your Notebook.
{ToDo] – How to share a Gemini Notebook with others, and make a publicly available notebook
Conclusion & Strategic Takeaways
Google’s Gemini Notebook represents an architectural shift in knowledge management: it unifies autonomous web research, grounded document interaction, cloud-based Python data computation, and multi-format content creation into a single, persistent workspace.
Account Tiers, Quotas, and Resource Limits
To manage system load, usage across Gemini Notebook is governed by compute allocations and tier structures:
- Free Tier (Standard): Supports up to 50 sources per notebook (with individual file caps of 500,000 words or 200MB). Daily allocations include 10 Deep Research queries per month, 3 Audio Overviews per day, 3 Video Overviews per day, and 10 daily studio generations (reports, quizzes, mind maps).
- AI Plus Tier ($4.99/month): Doubles general compute quotas and studio asset allowances.
- AI Pro Tier ($19.99/month): Quadruples baseline compute quotas, expands source capacity to 300 sources per notebook, grants 20 Deep Research queries per day, and provides access to advanced model reasoning.
- AI Ultra Tier ($99.99/month): Delivers the highest compute priority limits, maximum source capacities, and primary access to sandboxed Python cloud execution tools.
Operational Constraints & Health Warnings
- Grounding is Not Fact-Checking: Grounding restricts the AI’s source pool strictly to uploaded files, but it does not guarantee that the uploaded source files themselves are accurate. If uploaded documents contain errors or conflicting claims, the AI will reflect those errors. Users must click inline citations to independently verify claims against original source passages.
- Reset Mechanics: General compute allocations operate on a rolling ~5-hour reset cycle , whereas daily studio generation caps (such as Audio and Video Overviews) reset on 24-hour cycles.
- Isolated Compute Sandboxes: Python code execution takes place in isolated cloud environments. Only Python execution is supported, and code states do not persist permanently across separate user sessions.
As Gemini Notebook continues to bridge the gap between autonomous web research, native cloud computation, and multi-format content generation, the core friction of knowledge work is no longer gathering or formatting information. If Gemini Notebook can autonomously research the web, execute the math, draft the report inside your word processor, and publish the results as a video presentation, what part of the analytical workflow will you choose to keep for yourself?
Sources used in creating this blog post
Introduction to Gemini Notebook – published date: December 5, 2023
[TODO – more to come]