Mail

"I already have good research materials. But it takes too long to analyze and organize them."

If you've had that thought, this NotebookLM update is worth your attention. On June 8, Google integrated Gemini 3.5 and code execution into NotebookLM, transforming it from a passive document reader into an active data analysis agent.

What Changed — 3 Core Transformations
Code Execution: NotebookLM Now Does the Math Itself
Agentic Research: All You Need Is an Idea
11 Export Formats — From PDF to PowerPoint
Practical Use Cases for Educators and Researchers
What to Watch Out For: It's Not Open to Everyone Yet

1. What Changed — 3 Core Transformations

The old NotebookLM was powerful, but fundamentally a system that conversed based on materials you uploaded. You put in sources, and the AI answered from within them.

Three things changed after the June update:

Before	After
Answered only from uploaded documents	Combines external web search + internal sources
Generated only text and audio summaries	Can execute code and perform data calculations
Outputs: text, audio overview	Outputs: charts, PDFs, spreadsheets, PPT, and 9 more

Combined, these three shifts make NotebookLM feel less like a research assistant and more like a research workstation.

2. Code Execution: NotebookLM Now Does the Math Itself

From reasoning about numbers to actually computing them.

Previously, AI would "estimate" or "infer" numerical trends — calculations based on token prediction, which led to frequent errors. That's changed.

The June update gives each notebook its own isolated cloud virtual machine (VM). NotebookLM now executes Python-level data analysis directly:

Dataset normalization: Analyzes uploaded CSVs and auto-detects outliers
Statistical analysis: Runs mean, standard deviation, and regression analysis directly
Workflow automation: Instructions like "Extract KPIs from this file every week" are now possible
100+ pre-configured skills: Choose from sentiment analysis, text classification, keyword extraction, and more

NotebookLM code execution interface

"This isn't a calculator. It's closer to having a data scientist living inside your notebook."

3. Agentic Research: All You Need Is an Idea

A research agent that can start without any sources.

The biggest barrier to using the old NotebookLM was "you had to prepare good sources first." When you didn't know what to look for, it was hard to even begin.

Now it works in reverse. Start with a vague idea or question, and NotebookLM will:

Search Google directly for high-quality web sources
Find primary materials across multiple languages (English, Japanese, Spanish, etc.)
Expand the search to include recent work from related authors
Automatically add discovered sources to the notebook to build a knowledge base

The starting point for research has dropped from "having a finished source collection" to "having one question." A meaningful shift for researchers writing papers, teachers preparing lessons, and planners doing competitive analysis.

4. 11 Export Formats — From PDF to PowerPoint

Research done? Straight into presentation materials.

Old NotebookLM outputs were limited to text summaries and Audio Overviews. Useful, but you had to re-edit results in other tools.

The June update expands outputs to 11 formats:

Editable PDFs
PowerPoint presentations
Google Slides-compatible files
Spreadsheets (structured data)
Charts and infographics
Structured data files (JSON, CSV, etc.)

The Google Slides integration stands out. Generate a presentation draft in NotebookLM, then refine the design in Slides — with formatting intact, no broken layouts from the conversion.

5. Practical Use Cases for Educators and Researchers

As an EdTech CEO, what caught my attention most in this update is the potential to reduce lesson preparation time.

Teacher Scenario:

Start with "Find sources for a middle school climate change lesson next week"
NotebookLM automatically collects Ministry of Education resources, papers, and news
Generate PPT drafts for different student levels from the collected materials
Final polish in Google Slides

Researcher Scenario:

Start with a broad research question (e.g., "Status of AI literacy research among Korean teenagers")
Agentic research automatically searches for domestic and international related papers
Code execution runs statistical analysis on the collected data
Auto-output as charts and PDF report

Content Creator Scenario:

Topic research → draft generation → PPT/infographic output, all within one workflow

6. What to Watch Out For: It's Not Open to Everyone Yet

The limitations of this update must also be addressed clearly.

Access Restrictions: Code execution and agentic research are currently available only to Google AI Ultra subscribers and Workspace business customers (with AI Ultra Access or AI Expanded Access). Free users cannot access these features yet.

Gemini 3.5 Integration: The enhanced reasoning of Gemini 3.5 is included, but this model also only works under premium subscription conditions.

The global expansion timeline hasn't been announced, but considering Google's typical rollout patterns, it's likely to open to broader users within 6–12 months.

Closing Thoughts

NotebookLM's June update isn't just a feature addition. It's a paradigm shift in how research begins. You don't need finished sources. You don't need a precise question. Start with a vague idea and finish with analyzed data, presentation materials, and a PDF report — all within one tool.

AI tools are increasingly handling not just "search and organize," but "discover and analyze." NotebookLM is at the front of that change.

Related Posts

Sources:

NotebookLM June 2026 Upgrade: Gemini 3.5, Code Execution & Agentic Research

Table of Contents