NotebookLM: Your Externalized Brain, Without the Phantom Learning

Google keeps expanding its AI ecosystem, and strong synergies show up across certain tools, including NotebookLM, a very powerful research tool that benefits from the latest technological advances: the best AI models (Gemini), visual generation (photos with Nano Banana / videos with Veo) and even audio…

Here's an overview of Google's AI environment:

The idea is to take a full deep dive into NotebookLM, a very powerful yet still underused tool…

A refresher on NotebookLM

Just before sharing concrete examples, it's important to understand that the tool works differently from an AI model like ChatGPT or Claude.

The principle: you import your own sources (PDFs, Google Docs, websites, YouTube videos, audio files). The AI only answers from those sources. Never outside them.

In practice, this means far fewer hallucinations (if the information isn't in your sources, the AI doesn't make it up), and every answer is verifiable through a citation system that points to the exact source of the information.

Here's an example from the sources I shared:

So you decide which knowledge base the AI draws on. It's therefore not a generalist assistant, but rather a knowledge operating system. ChatGPT & co can share false information with you, and that's a problem when it's on a topic you don't fully master… because the errors are hard to spot.

You can also turn your sources into usable content (visual or not) thanks to the Studio.

For each space (“notebook”), you have 4 main sections:

1. The sources section to add and view all the sources used in your space
2. The discussion section where you find a summary of the sources and the history of your conversation. Suggested questions are available there.
3. The conversation section to ask your questions like on any model
4. The studio section to generate content and visuals from the sources. It's one of the most interesting parts.

Here, for example, is the variety of content types you can generate within the studio:

We won't go through every feature in detail, but I'm sharing 3 concrete usage examples. I give a warning at the end of the section about the possibility of phantom learning with the tool ⚠️

Going beyond the simple PDF summary

Summarizing sources is one of the first uses people have for NotebookLM; the idea is to go much further.

At Tandem, we regularly use the tool for specific cases: content production, internal training, competitive analyses…

So I'm sharing 3 use cases that tap into really interesting features and could replace your paid tools:

1️⃣ Create a sourced pitch deck without spending all day on it

2️⃣ Turn your messy notes into clean SOPs

3️⃣ Break down 8 competitors in 30 minutes flat

Spoiler: NotebookLM's real strength isn't the AI. It's the absence of hallucination. And in business, that changes absolutely everything.

Case 1️⃣: Pitch decks and strategic presentations

What it replaces: hours of research + slide generation in PPT + design in Canva.

Here are the 3 steps to follow:

1. Import the sources: Business plan (PDF), sector reports, competitor websites, financial benchmark, …

1. Choose slide generation and customize the instructions

1. The slides are ready in the studio with a complete structure and key messages per slide, verifiable citations for each data point and PowerPoint/PDF export (Google Slides coming soon)

Using deep research

In this example, you can also use the Deep Research feature, which adds sources from a prompt straight into your space.

The result: a list of source links for a specific need, ready for your deck.

Case 2️⃣: Turning notes into clean SOPs

What it replaces: scattered notes + hours spent aggregating and formalizing processes.

Example with employee onboarding documentation

You have scattered HR notes, recorded training videos, internal policies across different docs. The classic problem: everything is there, but nothing is structured.

You import everything into NotebookLM: onboarding call transcripts, Google Docs with the processes, internal videos, HR policy files, … and with a prompt like this in the "Rapport” section:

You'll be able to get an SOP, ready to be exported into your internal workspace.

The challenge of capturing project knowledge

It's a classic problem you run into in companies: a key contributor on a project leaves, and with them goes all the knowledge of a project or a process.

With NotebookLM, you can create one notebook per project or per critical process. You import all the existing sources into it (meeting notes, technical docs, explainer videos) to generate the missing documentation.

Onboarding a team member onto this project will inevitably be simpler: they access a centralized hub with all the knowledge formalized. No need to “ask X who knows how it works”.

Case 3️⃣: breaking down a list of competitors

You need to understand a competitive ecosystem: who the direct / indirect competitors are, value propositions, pricing, key features, user reviews, differentiation, acquisition channels, etc

The classic workflow with NotebookLM:

Step 1: Identifying the competitors (via Deep Research)

Step 2: Importing into NotebookLM

The competitor websites are imported directly into NotebookLM. The tool retrieves the relevant content (pricing, features, about, case studies pages).

Step 3: Extracting structured data

With a prompt and the Data table feature, you'll be able to structure all the information:

Step 4: Visualization

You can then move on to creating a market mapping (infographic, presentation, reports, mindmap) in a few clicks.

Here's an example of a comparative infographic of the competitive landscape.

Bonus: weekly industry monitoring in podcast format

Beyond one-off competitive analysis, NotebookLM is formidable for ongoing monitoring by combining: deep research + podcast generation

Deep Research lets you identify the 10-15 most relevant articles in your sector. You import them into a dedicated “Weekly monitoring” notebook.

Then, you generate a Audio Overview that runs 15 minutes and summarizes the week's key trends: podcast format, listenable in the car or on public transport.

The risk of phantom learning?

As you've seen, the tool can summarize hundreds of pages in a few seconds, generate quizzes, mind maps or even podcasts.

The problem? You feel like you're learning… when in fact you sometimes hand off all the cognitive effort to the AI.

One study shows that self-explanation (generative learning) improves performance by 0.55 standard deviations compared with passive reading, the equivalent of going from 12 to 16 out of 20. (source: Generative Learning Strategy (Self-Explanation))

In other words, real learning doesn't come from consuming summaries but from mental work: rephrasing, connecting ideas, explaining in your own words.

If NotebookLM spots the key ideas, frames the questions and structures the knowledge for you, you're optimizing consumption… not learning.

So be careful not to use the tool as an answer generator, but as a cognitive sparring partner.

In the prompt, you can prompt it to challenge your thinking, spot errors in your answers, etc