How I Used AI to Crack the Code of Viral Substack Notes (Part 2)

A 60-minute analysis that showed me why some Notes get 120 likes and others get 8 and how you can figure out your own patterns

Newsletter writers on Substack are often stuck in an endless trial-and-error loop with no feedback mechanism to understand what’s working.

Tell me you recognize this scenario: two nearly identical posts can get wildly different results, and you’ll have no idea why.

I posted this in the beginning of November:

“You don’t need to write huge essays. You need a rhythm. A pace your life can sustain.”

8 likes.

Two days earlier, I posted:

“A single post might do little. But 100 posts change everything. Consistency compounds like interest.”

49 likes. 5 comments. 2 restacks.

Same message. Same audience. Same time of day.

So what was different?

I stared at these two Notes, trying to find the magic ingredient. Was it the word count? The formatting? Some mysterious algorithm preference I didn’t understand?

Sure, I had some ideas. But with almost 20 years in research, I decided to use AI - Claude (you can use ChatGPT or any AI tool) - to scientifically investigate why some of my posts take off and others flop.

So I fed my Notes - successes and failures - into Claude and asked it to show me what I was missing.

What came back wasn’t a trick. It was a structural pattern that clearly showed me what I had overlooked.

In this article, I’ll share the simple analysis that helped me understand how to make my notes more visible and reach a bigger audience.

And how you can run the same analysis on your own Notes in under an hour.

If you’re not into the details, feel free to skip ahead to the blueprint and prompts below.

---

What This Article Will Give You

In the next 10 minutes, I’m going to show you:

✅ How to analyze your own Notes to identify what’s actually working

✅ How to decode successful competitors without copying them—understanding the structure behind their engagement

✅ The specific differences between Notes that flop and Notes that take off (with real examples)

✅ A 30-minute AI-powered analysis process you can run today to shorten your learning curve from six months to one afternoon

✅ How to apply these insights while staying completely authentic to your voice and message

This isn’t about becoming someone else. It’s about finally understanding the game you’re playing—so you can stop wasting creative energy on content that disappears and start building the audience you deserve.

Because here’s the truth: You’re not bad at this. You just can’t see your own blind spots.

Let me show you how to turn on the lights.

---

Step 1: I Analyzed My Own Best And Worst Notes (Playing Detective)

First, I gathered my Notes from the past few months and sorted them by engagement. I looked at likes, comments, restacks, and new subscribers.

I’ve analyzed my successful notes before, but this time I took a different approach and looked at both my best-performing and my worst-performing notes.

---

How I Used AI to Crack the Code of Viral Substack Notes (And You Can Too). Part 1.

---

Then I asked Claude (you can use ChatGPT or any AI tool) with this prompt:

“Here are my successful and unsuccessful Notes. Analyze them and tell me what factors determine why one works well but another gets just 5-7 likes. What patterns can you see?”

What the AI revealed about my top performers (More than 100 likes):

✅ Specificity + Vulnerability: My best posts included concrete numbers and milestones (514 subscribers, 7 months, “45% of writers quit”). My generic advice flopped.

✅ Community-Centric Language: The words which foster a sense of belonging, two-way communication, and deeper engagement between authors and subscribers work well.

✅ Meta-Content About the Platform: My Notes reflecting on Substack itself or writer statistics resonated hard. We want to understand the game we’re all playing.

✅ Actionable Insight + Personal Context: When I combined useful data (”45% quit within 6 months, but real progress happens after a year”) with encouragement, it hit differently than pure advice.

What tanked (8 likes):

My worst-performing Notes were abstract advices with no personal story, no community hook, and no specific numbers. It sounded like a generic motivational quote, not lived experience.

This quick analysis helped me figure out what to avoid in my notes if I want them to be noticed by others.

But you really understand things by comparing.

So, I decided to compare my most successful notes with those of another Substack writer who has a much bigger following - over 10,000 subscribers. Most of their notes get more than 100 likes.

---

Step 2: Analyze Your Successful Competitor (Learning from the Masters)

I collected 10 top-performing Notes of this author, fed them to Claude and asked it:

“Analyze the notes of my competitor. Tell me what are his strengths, why these notes take off, what patterns can you see?”

What emerged was really good.

First of all, Claude highlighted his strengths and derived this author’s strategy:

• Permission posts = High comments (people share their work)

• Milestone posts = Medium engagement (inspiration)

• Motivational posts = High restacks (shareable wisdom)

• Lifestyle posts = Moderate engagement (relatability)

Claude also explained why this strategy works at 10K subscribers and what practical takeaways can I apply for myself, given that I have just 780 subscribers.

“With a large audience, he can afford to be more directive and authoritative. His success story ($100K) gives him credibility to tell others what to do. Your gratitude-based approach works well for building community at your stage, but as you grow, mixing in some of his permission-giving + engagement-bait tactics could amplify reach.”

The key insight came when I asked Claude to compare this author’s most successful notes with mine!

---

AI Writing Tools for Newsletters: How to Create a Month of Content Ideas In 20 Minutes (Beginner-Friendly)

---

Step 3: Compare Your Notes to Theirs (The Mirror Test)

This is where it got uncomfortable and valuable.

I asked the AI to compare my approach to this author’s:

Other author/ My competitor

1. Action commands (”Promote,” “Start,” “Do it”)

2. Engagement bait (invites self-promotion)

3. Aspirational authority (shows income proof)

4. Short, punchy (3-5 sentences)

5. Repeats winning formulas verbatim

Me

1. Reflective observations (”What I love about Substack...”)

2. Gratitude statements (less interactive)

3. Relatable peer (celebrates milestones together)

4. Slightly longer, contemplative

5. Varies messaging each time

The insight that hit me:

I was writing reflections. He was building engagement infrastructure.

Neither approach is wrong. They serve different purposes. But if I wanted higher engagement, I needed to borrow some of such tactics.

Some writers worry that this kind of analysis might strip their notes of originality, feeling it’s like copying someone else’s work.

But this isn’t about copying someone else’s Notes.

It’s about understanding the system and what people genuinely resonate with, then applying those insights in your voice.

---

Step 4: What I Changed (Without Losing Myself)

Here’s what I started doing differently:

✅ Added more calls-to-action

✅ Included specific numbers

✅ Shortened my Notes

✅ Experimented with images

✅ Created repeatable “event” posts

Will see what results this will bring.

---

I Used AI to Analyze My Substack. Here’s the Growth Blueprint It Created

---

Your Turn: The 60-Minute Analysis Blueprint

Here’s how to do this yourself:

1. Gather your data (15 minutes):

• Collect your last 20 Notes

• Note the likes, comments, restacks, and new subscribers on each

2. Feed it to AI (5 minutes): Use Claude, ChatGPT, or Gemini with this prompt:

“Analyze these Notes and identify patterns in what performs well vs. poorly. What makes my top performers work?”

3. Analyze a competitor (20 minutes):

• Find 1-2-3 successful writers in your niche

• Collect their top 10 Notes

• Ask AI: “What patterns do you see in these successful Notes? What makes them engage their audience?”

4. Compare (10 minutes):

• Ask AI to compare your approach to theirs

• Identify 2-3 specific tactics you could adopt without losing your voice

5. Experiment (10 minutes):

• Write one Note using a new tactic

• Track the results

• Iterate

  ---

I Can’t Wait to Hear What You Discover

I’ve shown you my process. My patterns. My competitors. My insights.

But your data will tell a different story.

Maybe you’ll discover that your most engaged readers love your vulnerable notes, not your tactical ones.

Maybe you’ll find that your audience responds to questions, not declarations.

You won’t know until you look.

I’m genuinely curious. Because while I’ve shared my patterns, your patterns are probably completely different - and that’s exactly the point.

We’re not all trying to be someone.

We’re trying to be the best, most strategic version of ourselves.

And the only way to get there is to understand what’s already working, then do more of it.

So go run your analysis. Then come back and share with us what you discovered.

I’ll be here, reading every comment, learning from your patterns too.

Because that’s the real power of this approach: we’re all mirrors for each other.

Let’s crack this code together.

---