A growing share of shopping journeys now start with an AI agent, not a search box. A buyer asks ChatGPT for "a quiet office chair under $300 with free returns", and the model assembles a shortlist before the person clicks anything. Whether your product makes that shortlist depends less on your design and more on what the agent can read.
We call that property Agent Commerce Readiness: how well a store communicates its catalog to an automated buyer. It breaks down into three things.
What an AI buyer reads
When an agent evaluates a product page, it does not see your layout. It sees the machine-readable layer underneath:
- Structured data. JSON-LD
Productmarkup — name, price, availability, brand, ratings. This is the single strongest signal, because it is unambiguous. - Crawlability. Whether
robots.txtlets AI crawlers in, and whether you publish intent files likellms.txt. An agent that cannot fetch the page cannot recommend it. - Policy clarity. Shipping windows, return thresholds and warranty terms — stated as facts, not buried in prose a model has to guess at.
Why it decides the outcome
Two stores can look equally polished to a human and score very differently to an agent.
| Signal | Store A | Store B |
|---|---|---|
| JSON-LD price + availability | Present | Missing |
| AI crawlers allowed | Yes | Blocked |
| Return policy machine-readable | Yes | Buried in a PDF |
To a shopper both pages render fine. To an AI buyer, Store B barely exists — there is nothing reliable to quote, so the agent defaults to the competitor it can read.
The store with the clearest data wins the recommendation, not necessarily the store with the best product.
Where to start
You do not need a replatform. Most of the gap is structured data and a few crawlability fixes that ship in an afternoon. The fastest way to know your position is to measure it: run your store through Prefero and read the breakdown.
For the full scoring logic — what we crawl and how each category is weighted — see the methodology.