Run free scan
Menu
Playbook

What we found in the stores we checked

We checked 29 stores from 2 to 3 August 2026. We selected them for scanner testing; they were not a random sample. Six, or about one in five, were detected as WooCommerce. The numbers below describe only these stores on those dates. They cannot be used to describe online stores as a whole.

Benjamin Ong
Read the limit before the findings.This is a report on a small, chosen set. It is useful for seeing the kinds of problems our checker met. It is not a market study, a league table, or a claim about every store.
METHOD

How we counted

We used the latest stored check for each store. If a store had been checked more than once, it still counted once. The records came from earlier checker runs. The set included test candidates taken from published WooCommerce showcase lists. That choice shaped what we found.

Each domain counted as one store. We did not check whether more than one domain belonged to the same business.

Three stores returned only a homepage failure. The other 26 had deeper findings. A warning means our outside check could not settle the answer or found something that needed a person to review. It does not mean the store was broken.

ACCESS

Getting a clean first read was often hard

Nineteen of the 29 latest records had a homepage warning or failure. Nine were asked to slow down, six were refused, three failed to connect, and one returned a redirect that the checker did not accept.

Those results describe what our reader received at that moment. They do not prove that every AI tool was blocked. They do show why an outside check must report doubt instead of turning every failed request into a hard claim.

For the 26 stores with a public access-rules result, six allowed every reader we tracked. One explicitly blocked some tracked readers. Eighteen needed review because access was partial or the rules could not be read cleanly. One result did not support a pass or fail.

PRODUCTS

Finding a product page was the biggest limit

On 18 of the 26 deeper checks, the scanner could not locate a product page from the routes it tried. That is a discovery problem. It is not proof that the store had no product pages.

Five stores had a product record that was missing some useful facts. Two sampled product pages had no product record that software could read. One passed the full product-details check.

ANSWERS

Clear answers were rare in this set

The checker found no clear question-and-answer content on 24 of the 26 deeper checks. Two stores had question-style headings, but the check still needed work.

This does not tell us whether an assistant would mention a store. It tells us that the pages gave our checker little direct answer-style text to work with.

DISCOVERY FILE

The optional AI guide file was missing more often than present

Eight of the 26 deeper checks found a usable llms.txt file. Eighteen did not. That file can give a reader a short map of useful pages, but its absence does not make a store unreadable by itself.

THE LIMIT

What these findings do not say

We did not pick a random sample. We did not review every page on each store. We did not ask assistants to rank or recommend these stores. We did not measure visits or sales. Page speed was skipped in these stored runs, so this article gives no speed result.

The records are dated. A store may have changed since the check. The fair use of this article is to learn what can go wrong and what a careful check must leave open, then test your own store today.

Check the store you can act on

A broad average cannot tell you whether your own pages are readable. Run the free scan for a dated finding on your store, then check each warning before changing anything.