Something quietly changed in the last year. A growing share of purchases now starts with a question to an AI assistant instead of a search box. “What are the best wireless headphones under $200?” “Find me a meal kit that ships to Austin.” The assistant reads the web, compares options, and answers with a short list.
If your store is not on that list, you were never in the running. The customer never saw you.
Why AI search is different from Google search
Google ranks pages and shows the customer ten blue links. The customer does the comparing. AI assistants do the comparing for them, and they can only compare what they can read.
That is the key difference. An AI agent does not look at your hero image or your brand video. It reads structured data. If your price, availability, shipping policy, and reviews are locked inside images, JavaScript widgets, or vague copy, the agent skips you and recommends the competitor whose page it could parse.
The three things an agent needs from your store
1. Product schema on every page. Schema.org Product, Offer, and Review markup tells an agent exactly what you sell, what it costs, and whether it is in stock. This is the closest thing the agentic web has to a product feed, and most stores still ship pages without it.
2. An llms.txt file. Think of it as a catalog manifest written for machines: your products, prices, and policies in a plain, readable format at a predictable URL. Assistants are starting to check for it the same way crawlers check robots.txt.
3. Policies an agent can quote. Shipping costs, return windows, and warranty terms should exist as structured text, not a PDF or a support macro. When a customer asks “does this store take returns?”, the agent should be able to answer from your page directly.
What we do about it
Every page EPD Commerce publishes ships with all three by default. Schema markup is generated from your catalog, llms.txt stays in sync automatically, and policies render as structured data agents can read and quote. Support for the emerging agentic commerce protocols (ACP and UCP) is in beta, which means agents will eventually be able to transact, not just recommend.
The stores that win the next five years will not be the ones with the best ad budgets. They will be the ones AI can actually read.