I’m building a tool that sends an AI shopping agent through a store the way assistants like ChatGPT and Claude now shop for people — find a product, read the price, try the cart — and logs exactly where it gets stuck. It’s not launched yet, and I don’t want my own store to be the only thing it’s ever tested on. So the first 5 stores that reply get the full check free, with your permission — it never places an order and touches nothing in your admin. I’m the founder, happy to answer anything about how it works.
Yeah, I’d be interested in this. I think the part that’s easy to miss is that a store can work perfectly fine for a person but still break down when an AI agent is trying to navigate it. Variant selection, popups, weird cart behavior, apps injecting things, and all the little edge cases could probably cause problems that you would never notice yourself.
I’ve been using SiteGuru for the more traditional side of things, making sure important pages are crawlable, indexed properly, internally linked, and that there aren’t obvious technical issues. This feels like a different layer of testing that I haven’t really seen covered as much.
The logging is probably the part I’d find most useful. If you test enough stores, I’d be curious to see whether you start finding the same problems over and over, or if most of the issues end up being specific to certain Shopify apps and themes.
Sounds good - I can message you to gather any details. I am working through some coding right now so let me know and I can be in touch - Robert
Hi there @Robert_Dainowski
This seems like a good method to catch any storefront problems that might not surface during the regular testing cycle. I’d keep the test focused on the customer journey, discovering a product, selecting a variant, cart behavior, shipping, and checkout availability. You might also want to try loging out and use different devices, as sometimes the behavior of the storefront is different. A clean report indicating exactly which step the journey breaks on would make the results much easier to act on.
Thanks Roan, that is exactly the shape it has taken. The walk follows the full journey: find the product from a plain language request, read the product data, pick a variant, cart it, and check checkout availability. The report grades each step separately, so you see the exact step where an agent gets stuck instead of a pass/fail blob. It stops at checkout and never places a real order.
Good point on logged out state. The agent always walks the store cold, first time visitor, no account, which matches how shopping assistants actually hit a storefront today. On devices I will be straight with you: this is an agent’s eye walk, not device QA, so cross device rendering is out of scope. The grading rubric is public at Rubric — AI Store Shopper (Stelar Digital LLC) if you want to see exactly what gets scored.
Free checks are still open for anyone reading, and SealSubs stores are welcome too, Roan.