How we test and score
Last updated September 26, 2026
Every report combines two independent tests. One looks at what AI systems can read. The other looks at what an AI shopping agent actually lives through when it tries to buy.
1. Machine readability, scored out of 100
An automated scan of your public pages: robots.txt, your public product catalog and one product page. It never logs in and makes at most a handful of polite requests.
| What we check | Why it matters |
|---|---|
| AI crawlers allowed (OAI-SearchBot, PerplexityBot, Bingbot…) | Blocked bots can't show your products in AI answers |
| Product structured data: price, currency, stock, brand | What shopping engines read first |
| GTIN / UPC or MPN | Lets engines match and trust the exact item |
| Materials, sizing or dimensions, use case in product text | The questions shoppers ask assistants most |
| Shipping and return policy in structured data | "How fast" and "can I return it" decide the sale |
| Reviews exposed to machines, sitemap, SKUs, product types | Trust and discovery signals |
Each check has a fixed weight. Your score is the sum of the checks you pass.
2. Live AI shopper test, scored out of 45
An AI agent with its own browser, running from the United States, is given a realistic goal, such as "a queen-size sheet set" or "a men's shirt in size M". It opens your store fresh, searches, picks a variant, looks for stock, shipping and returns, adds to cart and stops at the checkout page. It never enters personal or payment details and never buys anything.
Nine steps are scored from 1 to 5 against fixed anchors, so two stores with the same experience get the same score:
| Step | 5 means | 1 means |
|---|---|---|
| Finding the product | Right product in 2 clicks or fewer | Could not find it |
| Price | Clear, next to Add to Cart | Hidden until cart or login |
| Choosing size or variant | One click, reflected in price and URL | Could not select |
| Stock status | Explicit "In stock" for the chosen variant | No signal at all |
| Product facts | Materials and fit or dimensions visible | Missing |
| Shipping | US cost and delivery time on the product page | Only after entering an address |
| Returns | Window, conditions and who pays, linked from the product | Not found |
| Path to checkout | No forced account, no surprise fees | Blocked |
| Blockers | No popup blocked a click | Two or more blocking popups or a CAPTCHA |
How findings are checked
Every report is reviewed before it is sent. Anything we can't reproduce on the live store is removed, and fixes are written in plain English.
Limits, honestly
We test public pages at one point in time, from one location, with one agent. Different AI systems behave differently and change often. A good score means fewer avoidable blockers, not a guaranteed ranking, recommendation or sale. No one can guarantee placement in AI answers.
See a sample report, with the store's name removed.