AI NativeWhy this matters now
When a shopper asks an assistant to find and buy something, the assistant acts outside its own knowledge. It reads structured data, picks stores it can transact with, and completes the purchase. If your store is hard to read or impossible to act on, the agent moves to a competitor that is easier. You never see the lost sale.
What an agent needs from your store
- A clean, current product feed: accurate price, availability, and specifications. Out-of-date feeds get filtered out.
- Product and Offer structured data on the page, so the agent can confirm what the feed claims.
- Stable product identifiers (GTIN, MPN) so the same product matches across sources.
- An action surface the agent can call: native UCP support on your platform, an actions manifest, or a Model Context Protocol endpoint.
- Crawl access for AI agents in your robots rules. Blocking the agent blocks the sale.
How to read your score
A readiness score is a starting point, not the goal. Passing every check means an agent can act on your store. It does not mean the agent prefers you. Preference is decided by feed quality, price, reviews, and how well your data answers the shopper's actual need.
That preference is what we measure as Share of Execution: of the buying questions where an agent acts, how often does it act on you rather than a competitor. Readiness gets you into the game. Share of Execution tells you whether you are winning it.
First moves
Fix the feed before anything else, because most stores lose on data quality, not on protocol support. Then add Product and Offer schema to the pages that matter. Then open an action surface. Re-check after each change so you can see which one moved your standing.