I asked AI to find me an e-bike and take it to checkout. Can you guess where it tripped?
Five US brands: Aventon, Himiway, Lectric, Ride1Up, Velotric. A normal task: a bike to ride to work, with a throttle, under $2,500 all-in, the cheapest one that fits. No payment. Just find it, add it to the cart, reach checkout.
I was sure it would hit the door. A captcha. A sign-up. "Prove you are not a robot."
It didn't. All five stores let it reach checkout.
Then it got interesting. At two stores you can't see the price with shipping until you give a full address. And the same request brought bikes from $995 to $1,999.
Then I asked what everyone asks before a big purchase: "What if it doesn't suit me after the first ride?" The product pages say "14-day returns" in big letters. The answer for a bike you have already ridden sits in the return policy.
Check your own site today:
- key features are in the product description, not only in reviews or filters;
- the total with shipping shows for a ZIP code, without a full address;
- the product page answers "what if it doesn't fit?"
So where did it trip? Not on bot protection.
At one brand the agent skipped a model that was $200 cheaper. It had a throttle. The store knew it: the tag was in the site's data. But the specs the agent read didn't say so. It found the throttle only in customer reviews, and picked the model where it was written plainly.
$200 on the receipt. Or a shopper who goes where things are written more clearly.
Letting AI onto your site is not enough. It has to find there what you already know about your product.
This is the first issue of the series "How to prepare an online store for shoppers' AI assistants". Free report, no sign-up: How to Prepare an Online Store for Shoppers' AI Assistants: Five US E-Bike Brands.
Want to see your site through an AI shopper's eyes? Send me a message.