I Spent 20 Years in Marketing. Then I Built an AI to Do What I Do.

I’ve spent 20 years doing one thing: figuring out why people buy stuff.

Not in a theoretical, MBA-textbook way. In a “here’s $50 million, make this liquor brand famous” way.

I scaled Fireball Whisky from distribution in one state to a billion-dollar global brand. I ran campaigns for Heineken, Hennessy, Buffalo Trace, and over 100 other brands across two decades.

Along the way, I noticed something that changed how I think about advertising forever.

The Pattern Problem

Early in my career, I was a creative. I believed in the magic of the perfect idea — the lightning bolt that strikes in a brainstorm and becomes a Super Bowl commercial.

Then I started looking at the data.

The best-performing ads weren’t the most creative. They weren’t the ones that won awards. They were the ones built on patterns — psychological structures that work across brands, across categories, across decades.

A great car ad and a great whisky ad and a great DTC supplement ad often share the same underlying architecture. Different words, different visuals, same psychology.

The problem? Finding these patterns manually is brutal work.

The 60-Hour Problem

Here’s what competitive ad research looked like at my agency, Emerald Digital:

1. Open Meta Ad Library
2. Search for a competitor
3. Scroll through hundreds of ads
4. Screenshot the ones that “look good” (completely subjective)
5. Try to remember why you saved them a week later
6. Repeat for 20+ competitors
7. Somehow synthesize all of this into a creative brief

That process took 40-60 hours per client. And the output was… a folder of screenshots and some vibes.

We were spending more time FINDING inspiration than CREATING from it.

The Run Time Revelation

The breakthrough came when I realized something hiding in plain sight:

An ad that’s been running for 6+ months is running because it’s profitable.

Not because someone forgot to turn it off. Not because the brand doesn’t test new creative. It’s running because the ROAS is too good to kill.

Run time is the single best proxy for ad performance that’s publicly visible.

Once I understood this, everything changed. I stopped looking at ads as individual creative pieces and started seeing them as data points in a performance landscape.

The ads that run the longest tell you what messaging works. The patterns across those long-runners tell you WHY it works. And the psychology behind those patterns tells you how to build your own version.

What I Saw Across 10,000 Ads

When we started systematically analyzing ads this way, clear patterns emerged:

Ridge Wallet — Every top-performing ad hits two notes: “slim” (identity) and “RFID-blocking” (fear). They don’t waste a single word on anything else. Their ads average 12 words. They’ve been running the same psychological structure for years because it works.

AG1 (Athletic Greens) — They run 150+ active ads at any given time, but they all fit into exactly 6 psychological angles: identity, fear, exclusivity, authority, social proof, and objection killing. Six angles, one product. Each angle reaches a completely different buyer.

HexClad — A cookware brand that nearly went bankrupt in 2019, now worth $500M+. Their secret? Fear-based messaging about forever chemicals in non-stick pans — backed by real news stories. They stack trust signals higher than anyone: Gordon Ramsay (who OWNS equity, not just endorses), “2M+ sold,” “60-day returns,” specific chemical testing data.

Liquid Death — $1.4 billion selling water. They’ve never once mentioned hydration, pH, minerals, or electrolytes. Their entire strategy is being the anti-brand in a boring category. They sell identity, not product.

Bellroy — The “quiet” DTC brand. No celebrity endorsements, no fear tactics. Just incredibly specific product demonstrations and a cult following built on craft. They prove you don’t need to be loud — you need to be resonant.

These aren’t random observations. They’re systematic patterns that repeat predictably across every successful brand we’ve studied.

So I Built a Tool

In 2025, I started building StealAds with one goal: make pattern recognition across ad creatives as fast and systematic as possible.

Not an ad spy tool that just shows you what competitors are running (Meta Ad Library already does that, poorly).

An analysis engine that:

• Monitors thousands of active ads across brands and categories
• Tracks run time as a performance proxy (longer = more profitable)
• Decodes the psychology — what emotional angle is each ad using? What trust signals? What copy structure?
• Surfaces patterns across brands that you’d never spot manually
• Gives you the frameworks to build your own winning creative based on proven structures

The idea: the brands spending $100M+ per year on ads have already done your testing for you. Their longest-running ads ARE the answer key. You just need a way to read it.

What StealAds Is NOT

Let me be clear about what we’re not building:

We’re not a “swipe and deploy” tool. Copying ads doesn’t work. The brands you’d copy from are already saturating their audience with that creative. By the time you clone it, the market has moved.

We’re not just a bigger ad library. More ads isn’t the answer. Better analysis of the RIGHT ads is.

We’re not replacing creative teams. Great creative still requires human insight, cultural awareness, and craft. We’re replacing the 60 hours of RESEARCH that happens before the creative work begins.

StealAds is for marketers who believe that pattern recognition beats brainstorming. That studying what works is more efficient than guessing. That the best ideas come from understanding psychology, not from staring at a blank brief.

The Bigger Bet

Here’s what I actually believe:

The future of marketing creative isn’t more AI-generated ads. It’s AI-analyzed patterns applied by human creative teams.

The AI is better at processing 10,000 ads and finding the signal. The human is better at taking that signal and making it culturally relevant, emotionally resonant, and brand-authentic.

StealAds sits at that intersection.

We handle the analysis. You handle the art.

What’s Next

We’re in early access right now, building with our first cohort of marketers and agencies.

If you’re spending money on ads (or planning to), and you’re tired of the screenshot-and-pray approach to creative strategy, I’d love to have you in.

Join the waitlist →

And if this resonated, I write about ad psychology, brand strategy, and what we’re learning from analyzing thousands of ads. Follow along — the best breakdowns are just getting started.

Matt Berman is the founder of StealAds and CEO of Emerald Digital, an AI-first marketing agency. Over 20 years, he’s scaled brands including Fireball Whisky, Heineken, and Hennessy. He lives in New Orleans with his wife Crystal.