Your next Amazon listing doesn’t need a perfect image. It needs to sell 100 units first.
A 7-figure Amazon reseller I know is walking me through what AI still gets wrong. His listings have gone from a month of a team’s work to a day. The main image is the one step that won’t shrink.
He sells electronics, and the AI keeps inventing parts. It adds a cable that doesn’t exist. It splits one that should be a single piece. It draws a rectangular device as a round one. It garbles the brand name on the box.
“You added four cables, there’s only two,” he says. “It’s simple letters. Why aren’t you just reproducing the letters that are on the photo?”
So his designer fixes it by hand. One main image can take her two or three days. His goal is 10 launches a day.
Then he catches himself.
“I’ve come to a realization,” he says. “I’m always shooting for perfect if I can. But perfect takes forever.”
His new plan: launch on the manufacturer’s images, get the listing live and the traffic coming, then have his designer replace the main image, because that’s the one shoppers click.
I tell him what we did when we sold equipment parts. We had tens of thousands of options, and most never sold. So every listing launched with one good-enough image and had to prove itself. Around 100 units a year earned a second image and better bullets. More sales earned video. Each upgrade, the listing paid for first.
“I take the same principle you just mentioned,” he says. “I call it sprinkle.”
Out of 2,000 listings, only 40 sit in FBA. When new products come out, his team sprinkles in four or five units of each. The ones that move get more stock.
He had the rule for his inventory. He just hadn’t pointed it at his images.
The 100-unit rule
In most catalogs, a small share of listings drives most of the sales. A perfect image on a listing that sells three units a year is effort you never get back.
- Launch on good enough. The manufacturer’s image is fine.
- Make it earn the upgrade. Around 100 units a year buys a better main image and better bullets.
- Keep earning. More sales buy more images, infographics and video.
When a listing does earn it, use AI for volume, not for one perfect shot. On an earlier call, I suggested he have Claude drive his image tool, generate 80 variations, score each one against his own checklist (cable count, shape, exact label text), and show him only the top eight. A person still makes the call.
Write the infographics for Amazon’s AI too. “Not compatible with X laptops” stops a return, and it answers a shopper’s question before they ask it.
If you’re an agency
The same math runs across every client launch. Put design hours where the sales already are, and say so in the plan, so the client knows why the slow SKUs launched on the manufacturer’s photo.
Earned, not given
The demos make AI images look finished. The gap between demo and production is a huge gap, and on Amazon it’s measured in cables, letters and returns.
Perfect still matters. It just has to be earned, 100 units at a time.
