How Gemini Notebook’s new agent mode turns it into a real research assistant

Gemini Notebook (formerly Notebook LM) has changed a lot. It’s no longer just a place where you upload PDFs and ask basic questions. With its new agent-style workflows, it can now research topics for you, analyze large datasets like YouTube analytics, and even turn insights into ready-to-use content frameworks.

From simple Q&A to real research assistant

Originally, Notebook LM was positioned as a “research assistant,” but in practice it mostly behaved like a smarter search box for your documents. You uploaded files, asked questions, and it quoted back what it found.

The new Gemini-powered version is much closer to a true assistant. Instead of just answering questions, it can be given a job: analyze data, find patterns, and generate useful outputs like scripts, frameworks, and creative assets.

Starting from a blank notebook

Previously, you had to begin every notebook by uploading sources. Now you can start from a blank notebook and let the agent build it for you.

When you open a new notebook, you’ll see a prompt like “What would you like this notebook to help you do?” From there, you can choose a preset like “I want to learn about a new topic.” Gemini Notebook will then ask what you’re interested in and offer to search the web or work with your own uploads.

For example, if you say you want to learn about scuba diving, Gemini Notebook will go out to the web, gather high-quality introductory material, and assemble a structured report for you. With a single click, you can import that research as sources into your notebook and continue exploring or creating from there.

Using Google Play Books as sources

One of the most interesting new features is Play Books integration. When you click “Add sources” in a notebook, you’ll now see a “Play Books” option.

This lets you connect the ebooks you’ve purchased through Google Play Books directly into Gemini Notebook. Once added, you can:

• Ask questions about the book’s content
• Generate summaries, outlines, or study notes
• Create new documents, ideas, or other assets inspired by those books

For readers who build their knowledge libraries in Play Books, this effectively turns your ebook collection into a searchable, generative knowledge base inside Gemini Notebook.

Connecting Gemini Notebook to your YouTube channel

Gemini Notebook also connects directly with YouTube, which opens up powerful analysis workflows for creators.

On your YouTube channel page, there’s an entry point labeled “Notebook LM” (which will likely be updated to “Gemini Notebook”). Clicking it lets you create a new notebook pre-loaded with your channel data. You can then rename this notebook, for example to something like “Channel analysis.”

Beyond that, you can export your lifetime YouTube metrics as CSV files (views, watch time, revenue, subscribers, and more) and upload them into the same notebook. By sorting sources by type, you can quickly separate analytics CSVs from transcripts or other documents.

Seeing Gemini’s chain-of-thought on your data

A big shift in the new Gemini Notebook is how transparently it shows its reasoning steps when working with your data.

Say you ask, “What’s my best performing video?” In the past, it might simply return a single answer. Now, it shows a visible chain-of-thought: which files it looked at, how it analyzed performance, and what metrics it considered.

For many channels, the “best” video might be an older one from a completely different content era. For example, a creator might have an early phase focused on Facebook or social media content, and a newer era focused on AI. You can refine the request by saying something like, “Exclude the social media era and focus on my AI-focused content.”

Gemini Notebook keeps both ideas in mind: the original “best performing video” question and the new constraint about which era to consider. It then responds with a far more nuanced breakdown.

Going beyond a single ‘best’ video

Instead of just naming one video, Gemini Notebook can segment performance across multiple dimensions, such as:

• Overall views leader
• Top video for revenue
• Strongest long-form engagement
• Best thumbnail performance
• Most efficient subscriber magnet (e.g., subscribers gained vs. views)

This multi-angle view helps creators understand not just what got the biggest numbers, but what actually worked for different goals—like attracting buyers, building watch time, or growing subscribers.

Automatic charts and visualizations

Gemini Notebook doesn’t stop at text summaries. It can also propose visualizations based on what it finds. For example, after analyzing your top AI-specific videos, it might ask:

“Would you like me to create a publication-quality chart comparing the performance of your top 10 AI-specific videos side by side?”

If you say yes, it will:

• Re-scan your analytics sources
• Select the relevant videos
• Generate a clean chart image showing metrics like total views, estimated revenue, and subscribers gained

The final chart is polished enough to drop into a presentation, report, or newsletter with minimal editing—something you might previously have expected from tools like Claude or ChatGPT, rather than from Notebook LM.

Turning Gemini Notebook into a YouTube content intelligence agent

The most impressive upgrade comes when you treat Gemini Notebook as an actual agent with a multi-step job, not just a Q&A bot.

For example, you can give it a three-phase assignment like:

1. Act as a YouTube content intelligence agent
2. Use the analytics CSVs and video transcripts in this notebook
3. Figure out what has actually worked on my channel and turn that into a practical framework for future videos

This kind of prompt forces Gemini Notebook to:

• Read large amounts of analytics data
• Compare strong vs. weak performers
• Study transcripts for patterns in hooks, pacing, and structure
• Hold all of that context while designing a reusable script framework

Finding strong and weak content patterns

In this agent-style workflow, Gemini Notebook first identifies your strongest and weakest content. It can detect:

• Top-performing long-form videos
• Underperforming videos that didn’t land
• Outliers, like Shorts that spike in views but don’t behave like your core content

Because it sees video length and format, it can recognize when something is a YouTube Short and treat it as a different kind of signal—what you might call a “Shorts mirage,” where views don’t translate into deeper engagement or subscribers.

Analyzing transcripts for what actually works

Next, Gemini Notebook dives into your transcripts and compares how your best and worst videos are structured. It looks at patterns such as:

• Types of hooks (belief-breaking hooks vs. simple scarcity pitches)
• How quickly the video sets up the problem and promise
• Time before the first real insight or “aha” moment
• Overall pacing and section flow
• Use of repeatable frameworks vs. one-off demos

This transcript-level analysis is where Gemini Notebook starts to feel like a real content strategist. It’s not just telling you which videos did well; it’s explaining why they worked.

Building a reusable script framework

In the final phase, Gemini Notebook turns those findings into a concrete framework for your future videos. A typical structure might include:

• Hook – A belief-breaking or curiosity-driven opening that quickly shows the viewer why this video matters
• System setup – A clear overview of the system, workflow, or framework you’re about to teach
• Quick win – An early, tangible result to keep viewers engaged
• Deep demo – A more detailed walkthrough that shows how to apply the system in practice
• Bridge – A transition that connects the demo back to the viewer’s broader goals or next steps
• Call to action – A specific, relevant next step (subscribe, watch a related video, download a resource, etc.)

Because the framework is tailored to your actual high-performing videos, it’s not just generic advice. It’s a distilled version of what has already worked for your audience.

From insights to ready-to-use scripts

Once the framework is in place, Gemini Notebook can go one step further and offer to generate a full script outline using that exact structure. For example, it might suggest:

“Would you like me to generate a 10-minute long-form script using this framework to outline your next video on source engineering inside Gemini Notebook?”

At that point, you’re only a few edits away from a fully developed, data-informed video script that’s aligned with what your audience already responds to.

Where Gemini Notebook fits among other AI tools

Many creators and professionals already use tools like ChatGPT or Claude for ideation and writing. What sets Gemini Notebook apart is how deeply it works with structured data and long-form sources inside a single workspace.

It can:

• Ingest large analytics files and transcripts
• Maintain long chains of reasoning over those sources
• Surface patterns and convert them into frameworks and assets
• Pull in external knowledge from the web or your Play Books library

If you’re interested in how agent-style workflows are reshaping tools, it’s worth comparing this evolution with other agent-focused systems, such as those discussed in why Odysseus agent mode is a real game changer for local AI. And for a broader view of how tools like Gemini are changing knowledge work, you may also find this piece on AI and the future of work useful.

Why this upgrade matters

The new Gemini Notebook shows what happens when a note-taking and research tool gains true agent capabilities. It’s no longer just summarizing documents—it’s:

• Doing targeted web research on your behalf
• Turning ebooks into living knowledge sources
• Acting as a YouTube analyst and strategist
• Designing frameworks and scripts that you can immediately use

If you work with large amounts of content or data—especially as a creator, educator, or analyst—Gemini Notebook’s new agent mode turns it from a passive reference tool into an active collaborator that can help you do real work.