Two purchases. That’s all I got after $202 in ads, 123 clicks, and a weekend of coding. Two people out of 37 who tried the free sample actually paid for AI-generated dating photos. The rest? They looked at their own face, smiled at the watermarked copy, and walked away.
If you’ve ever built something you thought was brilliant only to hear crickets, you know the feeling. It’s not just failure—it’s confusing failure. The numbers looked good. The tech worked. The price was reasonable. So why did almost everyone say no?
The product wasn’t the AI photo. It was the illusion of a match—and I sold the tool instead of the dream.
Let me walk you through the experiment. I had free Google Cloud credits, so I built a simple site: upload a selfie, get back cool AI-generated photos for dating apps. Three free with watermarks, then a $9.99 pack of 25. Backend did face recognition, pre-filtered scenes, good similarity. I launched Google Ads—8% CTR, $1.64 per click. Respectable metrics. Then reality hit.
The disconnect is brutal. The technical part—building an AI wrapper, training a face model, running ads—is now trivial. Claude can write the code. Google’s API can buy the traffic. But converting a curious clicker into a paying customer? That’s a psychological minefield.
You can’t sell a tool. You have to sell the outcome. Nobody wakes up wanting to buy AI-generated photos. They wake up wanting to get more matches, more dates, more validation. I gave them a hammer when they wanted a house.
One HN commenter nailed it: “Accept crypto, ideally Monero among others—who would want that tied to their identity?” That’s the real fear. Paying for “fake” photos feels shameful. It’s not just a transaction; it’s a confession. The customer isn’t buying an image—they’re buying plausible deniability. My funnel didn’t account for that.
The twist? I actually believed the product was good. And it was—technically. But in a world where anyone can spin up a similar site in 20 hours, the moat isn’t the code. It’s understanding the deep, irrational need underneath the surface.
So what’s next? I’m tempted to try Meta Ads. I like the idea of scraping competitor creatives that run for months—if they’re still up, they must be profitable. But I know now that better ads won’t fix a broken value proposition. The real question isn’t “How do I get more clicks?” It’s “Why would a stranger trust me with their self-image and their credit card?”
Building a functional AI product is no longer a competitive advantage. Understanding human psychology is. The $202 was a cheap lesson. The expensive one would be ignoring it.
FAQ
Q: Why did only 2 out of 37 free users convert?
A: Because the product solved a technical problem (generating photos) but not the emotional one (buying matches). Users felt shame paying for 'fake' images. The trust and perceived value gap was too wide.
Q: What's the practical lesson for AI founders?
A: Stop leading with your tech. Lead with the outcome your user desperately wants. Ask: 'What is the customer actually buying?' If it's not obvious without a manual, your funnel will bleed.
Q: Is advertising AI photo services viable at all?
A: Yes, but only if you reframe the offer. Sell privacy, anonymity, or a 'competitive edge'—not 'AI enhancement.' The successful players in this space hide the tech behind the promise of better dates.