My $10M VSL Script: How I Use AI to Write Video Sales Letters That Actually Convert

My name is Daniel Fazio. I make over $1 million per month, I'm the co-founder of Client Ascension and ListKit, and virtually everything I sell runs through a VSL funnel.

If you're not using VSL funnels, you're making a substantial mistake. But you're here now, so let me show you exactly how I build them using AI.

Who This Is For (And Who It's Not)

Before anything else: this process is exclusively for B2B offers. Specifically, offers where you are not selling to people making purely emotional decisions.

We're generating logic-based arguments, not emotional ones. If your market buys on emotion, this won't work. If your market buys on logic — ROI, results, process, proof — keep reading.

Step 1: Upload 10–15 Real Sales Call Transcripts

This is the core of the entire method.

You need 10 to 15 recordings of live sales calls from actual people who were interested in your offer. Not cold leads. Not fake demos. Real interested prospects.

A few important details on format:

• TXT format only — not PDF, not MP4, not MP3

• Extract the actual transcription file from your call recorder

• The best tools for this are Fireflies or Fathom — both integrate with Zoom and auto-generate transcripts

Once you have your transcripts, upload them to a Claude Project (you'll need Claude Pro for this — Sonnet 3.5 specifically). That's your project knowledge base. I personally uploaded 12 transcripts for one of my offers and could have gone up to 24 before hitting the context limit.

Step 2: Run the First Prompt — Extract Your Core Value Props

With your transcripts uploaded, start with this prompt:

"We have an offer where [describe your offer, who it's for, what it helps them accomplish]. I need your help writing a video sales letter for this offer. What are some core value propositions, key points, pain points, and questions I need to address in the video? Take time to contemplate this. If you need to ask me follow-up questions to generate a better answer, please do so."

This surfaces insights you might not have consciously noticed. It's also a gut-check on where your offer actually stands — what the market cares about versus what you think they care about.

Step 3: Extract Objections in Descending Frequency Order

This is the step most people miss, and it's what makes the whole thing work.

Run this prompt next:

"Can you extract the most common objections and questions, and list them out for me in descending order of the frequency with which they're asked?"

For a data offer I ran this on, the output looked like this:

1. Data quality and source questions

2. Volume and coverage

3. Price and value

4. Technical integration

5. Deliverability

6. Contract and commitment terms

7. Support and implementation

For a cold email setup offer, the output was:

1. Expected ROI and results

2. How the system works

3. Typical KPIs — leads and clients they can expect

4. What's included in the offer

5. Pricing

6. Case studies

Why Descending Frequency Order Is the Secret

Here's the math behind this.

If you have a large enough pool of prospects and you order objections by how frequently they come up, that order naturally reflects the logical sequence those questions get asked in. Each answer triggers the next question.

You answer question one → that answer raises question two → you answer question two → that raises question three. And so on.

In a VSL, your job is to anticipate each question before the viewer asks it. That's the difference between marketing and sales: marketing is a sales call without the other person present.

A VSL is literally that. It's you running a sales call on video. So the smartest thing you can do is extract the exact questions your real prospects already ask — in the exact order they ask them — and answer them sequentially.

That's the entire framework.

Your VSL Table of Contents Writes Itself

Once you have the objection list, your VSL structure is basically done.

For the cold email setup offer, the table of contents maps directly:

• Section 1: What kind of ROI can you expect?

• Section 2: Here's exactly how the system works

• Section 3: The KPIs and what's realistic

• Section 4: What's included

• Section 5: Pricing and what you get

• Section 6: Proof — client results and case studies

It's not creative. It's not clever. It's just answering the questions your prospects already have, in the order they already have them, using data pulled directly from your own sales calls.

That's it. That's the method that's generated over $10 million.

The Takeaway

Most people try to write VSLs from scratch — guessing at what matters, what order to say it in, what objections to handle. That's why most VSLs don't convert.

The smarter move: let your real sales conversations do the work. Upload the transcripts, let Claude extract the patterns, and let the data tell you exactly what to say and in what order to say it.

Marketing is just a sales call without the other person present. Build your VSL like one.