A seller came to me with 47 AI-generated product images. None of them were worki
By Kamaljit Singh · June 3, 2026 · Curated by George's Blog
A seller came to me with 47 AI-generated product images.
None of them were working.
The images weren't bad. They were actually stunning. Clean backgrounds, perfect lighting, lifestyle shots that looked like a $10k photo shoot.
But conversions were flat. CTR was dying. And he had no idea why.
So I asked him one question:
"Where did the concepts come from?"
He looked at me blankly.
"I just... described the product and told it to make something good."
There it was.
47 images built on vibes. Zero images built on data.
No customer reviews. No competitor gaps. No buyer psychology. No understanding of what his specific customer was afraid of, hoping for, or needed to see before they'd click.
Just prompts. And hope.
That's the dirty secret of AI image generation right now. It's incredibly good at execution and completely blind to strategy.
So that's what we built.
Listing Optimization AI pulls the real signals first.
Reviews, objections, market gaps, purchase triggers and builds every image concept around that data.
Then it shows you the reasoning. Right next to every image, you can see exactly which insight drove the creative decision.
Not "here's a nice image."
"Here's an image designed to overcome this objection, sourced from these reviews, targeting this moment in your buyer's decision."
The seller relaunched with 6 images.
47 → 6. Vibes → data.
That's the difference.