Is your online store ready for AI search?

More and more people no longer type their query into Google, they ask an AI assistant. It reads stores, compares products and recommends one. So the decisive question for your online store is no longer just “do I rank on Google”, but: can an AI even understand my store?

What “AI search” means for a store

ChatGPT, Claude and Gemini no longer just answer questions, they actively browse. A user writes “find me a waterproof motorcycle jacket under 200 euros”, and the assistant opens stores, reads product pages and presents a recommendation. That is exactly where it is decided whether your store shows up or stays invisible. And assistants read differently than people: no eye for nice design, just the raw structure underneath.

If your product page is really a wall of JavaScript that only assembles inside the browser, with no machine-readable data, the AI often sees only fragments. Price, availability, variant, all blurry. It moves on to the next store that makes it easier.

There is even a score for it

Google Lighthouse, the standard tool behind PageSpeed Insights, has an experimental category called Agentic Browsing. It rates how well an AI agent can read and operate your page. I measure this regularly on real stores, and the numbers are sobering: many classic WooCommerce and Shopware stores sit below 30 out of 100 on desktop. The same stores often score 90 or more on SEO. In other words: found by Google, missed by the AI.

Why most stores fail

• No product schema. Without structured data (JSON-LD) the AI has to guess price and availability from body text instead of reading them cleanly.
• Everything depends on JavaScript. If the content only renders in the browser, many agents see nothing but an empty shell.
• Too slow. If the page needs seven or eight seconds to show content on mobile, an agent bails out just like an impatient customer.

Test it yourself in five minutes

You do not need special knowledge to get a feel for it:

• Run a product page through PageSpeed Insights and look at load time and layout stability.
• Use the Schema validator to check whether your product pages expose a Product schema with price and availability at all.
• Just open your domain plus /llms.txt. If you get a 404, no assistant knows what to do with your store.

The levers that actually matter

Structured data. A clean Product schema with name, price, availability and reviews is the language machines use to understand your catalogue. It is the single biggest lever and usually done in days, not weeks.

Serve the content server-side. When price, description and availability are already in the delivered HTML instead of appearing later in the browser, every agent reads them reliably. As a side effect the page also gets faster for real customers.

Clean semantics and an llms.txt. Proper headings, real alt text, clear buttons. Plus an llms.txt that tells AI systems in one short file what your store offers and where the important pages are.

The endgame: headless

If you want the maximum, you decouple the storefront from the admin. The store stays in the background, the frontend is rebuilt and delivered blazing fast. Load times under a second and top scores in agentic browsing become normal rather than the exception. A real example is alleswurst.at: PageSpeed 100, under a second to content, equally readable for human and machine. How such a rebuild works is explained under Headless WooCommerce.

AI search will not replace Google overnight, but it is becoming a second entrance to your store. Keep it open and you sell to customers you would otherwise never have seen. If you want to know where your store stands right now: headless.markusstoeger.com measures it for free. Or write me directly and I'll tell you which lever pays off most for you.