Make your store readable to AI shopping agents
More shoppers are asking ChatGPT, Claude or Perplexity to find and buy things for them. When an agent hits your store, it behaves nothing like a browser: it never sees your theme, it does not scroll, and it has no patience for a captcha. Either your catalog answers it in minutes, or the agent answers from a competitor who does. Here is what actually decides which one happens, in the order it matters.
1. Let the right bots in
Start with robots.txt, the front gate. The crawlers that feed shopping answers - GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot - all obey it. Block them and your products quietly disappear from the answers their models give, no matter how good your catalog is.
User-agent: GPTBot Allow: / User-agent: OAI-SearchBot Allow: / User-agent: PerplexityBot Allow: / User-agent: ClaudeBot Allow: /
Two nuances worth knowing. First, user-triggered agents (ChatGPT-User, Perplexity-User, Claude-User) ignore robots.txt by design - they act as the shopper's own browser, so robots.txt cannot stop them anyway. Second, if you run a bot manager or WAF, check it is not tarpitting those same user agents: a wall that hangs the request reads to the agent exactly like being blocked. That is the single most common thing our free check finds.
2. Publish machine-readable products
A screenshot of your homepage is not a catalog. An agent needs products it can parse: names, prices, stock, a link, an image. If you are on Shopify, /products.json already exists on your store - test it. On WooCommerce, the store API does the same job. On anything custom, schema.org Product JSON-LD on every product page plus a clean sitemap gets you there.
Keep the details honest: price including the rules that change it, in-stock or not, and product images at 800 pixels wide or more - most agent specs require it, and a 200-pixel thumbnail is how a product ends up unrepresentable.
3. Give them a route to the cart
Reading is not selling. An agent that has picked the product still has to complete the purchase, and most stores give it no path: the cart is a JavaScript session, the checkout demands an account first, or the buy button simply is not in the data the agent can see. What works is a plain cart URL the agent can construct and hand back to the shopper - prefilled with the product and quantity, landing on your own checkout. Shopify stores already have one; the pattern is /cart/[variant_id]:[qty]. If your platform does not, that gap is where you lose the sale after winning the recommendation.
Your three options, honestly
Everything above you can do yourself, free, in an afternoon - and if you have a developer, that is a reasonable route. The alternatives: do nothing and hope the captcha keeps the traffic out (it keeps the agents out, and the agents carry the demand); or join a platform's shared catalog, where the platform lists your products inside channels it controls, sets the conversation, and keeps the data. The trade-off is ownership: the platform's agent answers from your catalog, and you learn nothing about what shoppers asked for.
The third route is running the front door yourself: the same machine layer - the manifest, the product answers, the cart links - on your own domain, with every agent question recorded so you can see why someone's agent wanted "trail shoes, a gift for my sister". That is what Clair's free check looks at first: it visits your store the way a shopping agent would and shows you exactly where the journey breaks. Owning the front door is the paid half, $299 a month per store.
Clair itself is run end to end by AI agents on NanoCorp, which is why this guide stays current as the agent specs move.