The AI OS · Letter #26
March 15, 2025

I Stopped Trusting AI Blindly—Here’s My Research Hack

Perplexity + NotebookLM = a research superpower. Here’s how to use them together.

I Stopped Trusting AI Blindly—Here’s My Research Hack

Hey friend,

Many of us rely on AI to answer quick questions.

I ask, I get an answer. Simple.

But what happens when you need more than just an answer? When you’re researching a topic, collecting insights from multiple sources, and trying to piece together a well-structured, deep analysis?

That’s where things get messy.

Last week, I wrote about Deep Research, an AI-driven method where a model searches the internet, reflects on its findings, and generates a fully referenced, well-crafted report on a subject. It’s a solid solution, but it also has some limitations:

In short: powerful, but expensive, slow, and not easily personalizable.

I tested every solution under the sun. The only workflow that truly worked for me?

→ Perplexity + NotebookLM.

(You can replace “Perplexity” with any AI search engine—ChatGPT, Grok, etc.)

(If you're not familiar with these tools, check out my previous letters about using AI for search, Perplexity, and my NotebookLM crash course)

Used together, they create a research superpower.

In this newsletter, I'll show you six  ways to use Perplexity and NotebookLM to level up your research, learning, and content creation.

Whether you're a marketer, entrepreneur, content creator, or just an insatiably curious human—this workflow will save you time, reduce AI hallucinations, and help you extract better insights.

Let's dive in.

Why Use Both Perplexity and NotebookLM?

Think of Perplexity as your AI-powered researcher—fetching up-to-the-minute web data, finding reports, studies, and articles in seconds.

But there's a catch.

Every time you search, Perplexity starts fresh. Different sources, different results. Even if you upload your own documents to Perplexity Spaces, it still leans on its “pre-trained data” and system prompts.

NotebookLM, on the other hand, is like your AI-powered analyst—working only with the sources you provide. It won’t hallucinate facts or pull in random information.

In a nutshell:

That means you can:

✅ Use Perplexity to gather quality sources.

✅ Import those sources into NotebookLM for deep, structured analysis.

✅ Ask questions, extract patterns, and organize knowledge—without relying too much on LLM BS.

Now, let's talk about how to use this combo in real-world scenarios.

1. Market Trend Analysis

🔍 Find emerging trends & craft thought leadership content.

As a marketer, staying ahead of industry trends is a full-time job.

Perplexity can help automate part of the process:

1️⃣ Create a Perplexity Space named "[Your Industry] Market Research."

2️⃣ Set up custom instructions—for example:

3️⃣ Perplexity will now auto-fetch relevant sources whenever you trigger a search.

4️⃣ Manually review & import high-value sources into NotebookLM.

5️⃣ Ask NotebookLM:

🎯 Takeaway:

2. Product Research & Enhancement

💡 Use real customer insights to improve your product.

Let’s say you're building an email marketing automation tool. You need to know:

Step 1: Find Customer Feedback with Perplexity

Step 2: Analyze Competitor Weaknesses with NotebookLM

Import review sources and ask:

🔥 Example:

If reviews say "Lemlist lacks reporting capabilities" and "has problems exporting leads as CSV," that's your opportunity to differentiate.

🎯 Takeaway:

Some might say this can be done using deep search or search functionalities alone. True, but NotebookLM offers superior analysis capabilities, lets you combine sources from your computer (with the option to focus on specific sources), and is specifically optimized to avoid hallucinations. It can also suggest FAQs, generate briefings, save your notes, and much more (cf. my previous letter). And it's free.

3. Deep Audience Research

🎯 Understand your target market.

Imagine you run a childcare business and need to market to modern parents.

Step 1: Find Audience Insights with Perplexity

Step 2: Extract Key Insights with NotebookLM

Import your sources and ask:

Step 3: Refine Your Business Messaging

Import your website copy and ask:

🔥 Example:

If NotebookLM highlights that parents prioritize flexible scheduling, but your website doesn’t mention it, that’s a problem worth fixing.

🎯 Takeaway:

I did this exact same exercise for my company’s website. I searched social media and sources on insights about learning AI and using it for business performance (I activated “Deep Search” to go deeper).

Prompt: "Search specifically for:

  1. Individual user experiences and challenges when learning to use AI tools in daily work (focus on common frustrations, learning curve issues, and implementation barriers)
  2. Case studies and reports documenting startup and enterprise-level AI integration challenges, with emphasis on medium organizations (50+ employees)Include recent discussions from professional forums, industry surveys, and verified user testimonials from the past 12 months."

I imported the Perplexity results into NotebookLM, added my website as a source, and asked it to suggest improvements.

Key takeaway: we need to talk about prompt engineering very quickly on the page...

4. Podcast Research & Positioning

🎙️ Plan a high-value podcast based on real audience preferences.

Let’s say you’re launching a music podcast.

Step 1: Find Top-Performing Podcasts with Perplexity

Step 2: Extract Insights with NotebookLM

🔥 Example:

If users complain that "most music podcasts are too theoretical," you might position yours as practical & no-nonsense.

🎯 Takeaway:

5. Learning Any Subject Faster

📚 Use AI to accelerate your learning process.

Say you want to self-study Marketing Psychology.

Step 1: Find Quality Learning Materials

Step 2: Turn It Into a Custom Course with NotebookLM

🔥 Example:

You learn about anchoring bias (how the first price a customer sees shapes their perception of value). You ask NotebookLM:

"How can I apply anchoring bias to my childcare business?"

And it suggests:

🎯 Takeaway:

6. Improving Public Speaking & Presentations

🎤 The art of delivering engaging speeches.

Step 1: Find Great Speaking Examples

Step 2: Analyze Speaking Techniques with NotebookLM

🔥 Example:

If top speakers use personal anecdotes to open, you might start your next presentation with:

"Back in 2015, I made a huge marketing mistake that cost my company $50,000. Here’s what I learned..."

🎯 Takeaway:

Here's a helpful tip: You can use NotebookLM to make your presentations better. Just upload your slides and script, then ask it to suggest improvements based on the successful speeches you found!

Final Thoughts

By now, you can see how powerful the combination "AI-Search × NotebookLM" can be.

I use this exact workflow for everything—from researching and analyzing topics to crafting video scripts, enhancing presentations, and polishing copy.

The possibilities are bounded only by your imagination.

Try this out, experiment with different prompts, and share your findings!

And if you found this helpful, check out my AI Bootcamp if you want to go further in your AI journey and book a spot for the next cohort (subscription ends on April 30th). Some of you couldn't book a spot last week—spots are filling up fast. Don't worry; just book a spot and we'll get back to you quickly, as we're adding more spots.

Thanks for reading. Until the next one,

—Charafeddine

New letters now publish on charafeddine.co

Read the latest letters