Learn
How to check if ChatGPT can see your products
Asking ChatGPT whether it can see your store is the least reliable way to find out, because an assistant answering from memory produces the same confident sentence whether or not it fetched anything. What decides the answer is mechanical: whether your robots and CDN rules permit the fetch, whether your catalogue can be enumerated, and whether your product data is readable once retrieved.
Asking ChatGPT whether it can see your store is the least reliable way to find out. An assistant answering from what it already knows produces a confident, plausible description whether or not it fetched a single page from you, and nothing in the reply distinguishes the two.
The answer is decided by three mechanical things, and all three can be checked directly.
One: are you permitting the fetch
Check your robots.txt for rules that catch assistant user agents. The usual accident is a disallow written to keep content out of training data that also removes you from assistant driven shopping, because both were listed together in whatever guide was followed. Those are separate decisions. See robots blocks agent.
Then check the layer above, at your CDN, where bot categories are usually set once and forgotten. This setting changes without anybody editing your site.
Then check that a well behaved agent is not being challenged. A JavaScript interstitial is unanswerable, so it makes every other question moot. See bot wall.
Two: can your catalogue be enumerated
Being permitted to fetch is not the same as being findable. Confirm your sitemap lists product URLs rather than stopping at collections, and that a product feed, if you publish one, resolves and is current. See no product feed.
Three: is your data readable when it arrives
Open a product page with JavaScript disabled and read it. The price, the availability, the variant options and the terms all need to be present as words or structured data. If the price only appears once scripts have run, an assistant reading your markup has no price. See missing price data.
What a good test looks like
If you do want to try the assistant itself, make the question one that can only be answered from a live fetch. Ask about something that changed recently and is not in any cached copy: a product you added this week, a price you moved yesterday. A correct answer to that is evidence. A correct answer about your flagship product is not.
Even then you have tested visibility, not purchasability. Where a purchase would complete, and what breaks on the way, is covered in can ChatGPT buy from your store.
Questions
Why can it describe my store if it cannot see it?
Because it may be answering from training data rather than from a live fetch, and the two are indistinguishable in the reply. A confident description of your store is evidence that your store was written about somewhere, not that your catalogue is readable today.
It found my product but got the price wrong. What does that mean?
Usually that it read a stale source rather than your live page: an old cached copy, a feed that has not refreshed, or a third party listing. That is a freshness problem, and it is worth fixing, because a wrong price loses comparisons in both directions.
Should I test with a private window or an account?
Neither changes what the assistant can fetch, which is the thing you are testing. If you want a test that is actually about your store rather than about the assistant, check the fetch permissions and the data directly.