Your Store Is ‘Agent Ready’ by Default, We Tested 122 Shopify Stores to See What That Actually Means

There is a dangerous gap between a website that can be read by an AI and a store that can actually be used by an AI agent to complete a transaction. Most of us assume that if the technical plumbing is there, the customer experience follows. But in ecommerce, the "plumbing" is only half the battle. This connects with No AI Agent Reads It Yet when the same signal needs a clearer operating decision. The same pattern also shows up in Beyond Brand Sovereignty, where the practical question is how the signal becomes visible.

I've been thinking about the difference between a site that acts as a brochure and one that acts as a storefront. A brochure tells you what exists; a storefront lets you buy it. When we move into the era of AI agents, this distinction becomes critical. If an agent can find your product but cannot answer a basic question about your return window or shipping costs, the transaction dies before it ever reaches the checkout.

The Gap Between Platform Rollouts and Merchant Readiness

Recently, Slobodan Manic pointed out that the text only versions of sites being served to AI often strip out the very things a user needs to actually take action. He suggested a three part framework for ecommerce teams: expose what can be done, make it callable, and report the outcome. This prompted a deeper look into how the largest ecommerce platform in the world, Shopify, is handling this. A useful companion note is Working Framework, because it looks at a nearby part of the same system.

The data shows a clear trend: Shopify has done the heavy lifting, but the merchants haven't. In a test of 122 Shopify stores, the vast majority were using the same Shopify generated templates and scripts. Specifically, 89 of 122 stores served an agents.md file, but almost all of them were identical copies of a Shopify template. Only five stores in the entire sample had written their own.

This means that if you see a store that looks "agent ready," you aren't necessarily seeing a forward thinking merchant. You are seeing Shopify's rollout schedule. The platform has enabled the tools, but the merchants haven't customized the logic.

Expert Interpretation: The tradeoff here is convenience versus precision. Shopify provides a "turnkey" AI readiness, which is great for getting started, but it creates a false sense of security. Merchants might believe they are optimized for AI search and agents when, in reality, they are just using a generic template that lacks the specific commercial rules needed to close a sale.

The Scale of Shopify's Agent Infrastructure

It is important to acknowledge that Shopify has achieved something massive here. No other platform has deployed an agent callable commerce surface at this scale. By implementing the Universal Commerce Protocol (UCP) and MCP tool listings, they've given millions of merchants instant access to catalog search, cart management, and checkout tools without requiring a single line of custom code.

The technical execution is surprisingly strong. In tests involving creating checkouts for in stock items, reading them back, and canceling them, the system performed well. Unlike traditional web forms that might just show a "success" message for human eyes, Shopify's tool layer provides machine readable errors. If an agent tries to cancel an already canceled order, the system refuses the request with a clear error rather than silently repeating the action.

Expert Interpretation: This is a significant win for the "callable" part of the equation. The infrastructure is stable and predictable. However, the decision maker should realize that technical stability is not the same as a conversion optimized experience. A fast API response doesn't matter if the agent doesn't know the shipping policy.

Callable vs. Completable: The Critical Distinction

There is a nuance in how agents enter the checkout process. There are two main routes: a server endpoint (UCP over MCP) and browser based tools (WebMCP). While the server route is highly efficient, with median response times around 1.1 seconds, there is a fundamental ceiling on what the agent can actually do.

In every test conducted, the checkouts returned a status of requires_escalation. This is a deliberate safety feature in Shopify's default settings: an agent can build a priced cart, but a human must authorize the final payment. This is a reasonable guardrail, but it redefines what "agent ready" means. Currently, the agent is an assistant that prepares the order, not a proxy that spends the money.

there is a difference between an agent being able to "call" a function and being able to "complete" a purchase decision. A purchase decision requires rules. For example, if a store adds a "Free Returns Coverage" fee to a cart automatically, the agent can see the charge, but it often lacks the rules to explain why it's there or how to remove it. The agent can execute the action, but it cannot navigate the commercial logic.

Expert Interpretation: The tradeoff here is security versus autonomy. By requiring human intervention for payment, Shopify protects the merchant and the buyer. But for the merchant, this means the "last mile" of the AI experience is still a manual one. The goal should be to make the transition from agent to human as frictionless as possible by providing all the necessary context upfront.

The 'Ready By Default' Trap

This is the most important point for ecommerce leaders: do not mistake a passing score on a technical readiness checker for a functional AI shopping experience. A store can have an agents.md file and a UCP profile and still fail the moment a buyer asks a real question.

When stores were tested with questions a real buyer would ask, the results were poor. While return windows were generally handled well, other queries failed. For instance, a store might have a clear policy stating "free standard ground shipping on U.S. orders over $75" in their FAQ, but the AI agent fails to retrieve that specific answer when asked in a natural, buyer centric way.

The data is stark: across nine checks used to determine if an agent could fully resolve a purchase decision, zero of the 122 stores passed perfectly. The top performers only reached seven of nine, usually failing because they had no stated cancellation window or no way for the agent to trigger a cancellation.

Expert Interpretation: This is a "semantic floor" problem. The information exists on the page, but it isn't structured in a way that the agent can reliably retrieve it as a value. Merchants need to move beyond "having a policy page" to "having machine readable rules."

An Organizational Problem, Not a Technical One

Interestingly, the failures aren't usually the fault of the web team or the SEO team. The things the stores pass are the automated platform features: pricing, variants, and checkout. The things they fail are commercial rules, like cancellation windows and shipping nuances.

These rules are typically owned by operations, logistics, or finance teams, not the people managing the website. This is an organizational gap. The search team cannot fix a missing cancellation policy if the operations team hasn't defined one or communicated it to the digital team.

Expert Interpretation: To solve this, ecommerce leaders must treat "Agent Readiness" as a cross functional project. It requires a bridge between the people who set the business rules and the people who manage the digital storefront. If the rule doesn't exist in a structured format, the AI cannot "hallucinate" a correct one.

How Shopify Can Close the Gap

Because Shopify operates at such a massive scale, they have the opportunity to fix these gaps globally. Much of the missing data already exists within Shopify's own admin settings in structured form. For example, Shopify already tracks return windows (14, 30, 90 days), restocking fees, and cancellation windows for unfulfilled items.

If Shopify were to automatically push these structured settings into the agents.md files and the policy tools, the majority of the "decision gaps" would vanish in a single update. passing delivery dates directly through the protocol would provide the immediate clarity buyers need before committing to a purchase.

Expert Interpretation: The opportunity here is to move from "providing the tools" to "providing the data." The infrastructure is ready; the content is the bottleneck. Any platform that can automate the translation of admin settings into agent readable rules will win the AI commerce race.

Practical Steps for Merchants: Verify and Refine

You don't need a massive audit to see where you stand. You can verify your store's agent readiness in an afternoon by simply asking your own policy tools the questions your customers ask.

Test your store for the following:

The specific return window (in days). Exact shipping costs for a specific address. Estimated arrival times. Whether local pickup is available today. The policy for open box or used returns.

If the agent returns a link to a page instead of a direct answer, or if it gives an empty response, that is exactly what your customers are experiencing. To fix this, you should:

Set your return and cancellation rules clearly in the Shopify admin. Write these same values as plain numbers in your FAQ and policy pages. Enable delivery dates if your logistics can support them. Check your robots.txt to ensure you aren't accidentally blocking AI from reading your /policies/ folder. Extend your default agents.md file to include your actual commercial rules.

Expert Interpretation: The goal is to reduce the "cognitive load" for the agent. The more you can provide values (e.g., "30 days") instead of prose (e.g., "We offer a generous window for returns"), the more likely the agent is to successfully guide a user to the checkout.

Moving Beyond the Brochure

A text only version of a website is essentially a brochure. Shopify has provided the "checkout" part of the experience, but the "rulebook" is still missing. Having a working checkout is useless if the agent cannot answer the basic questions that lead a customer to use it.

An agent needs four things to resolve a purchase: information, context, rules, and operational capabilities. Most stores today have the operational capabilities (the checkout) and some information (the catalog), but they are missing the rules and context. Until those are filled in, most stores are only "agent ready" until the buyer asks a real question.

One final thought: most of these tests assume the buyer already knows exactly which product variant they want. In reality, the hardest part of the journey is the decision logic, choosing between sizes, models, or related products. That logic is entirely up to the merchant to define and expose.

Expert Interpretation: The ultimate competitive advantage in AI commerce won't be who has the best API, but who has the most transparent and accessible business rules. The stores that win will be the ones that make it easiest for an AI to say "Yes, this is the right product for you, and here is exactly how the shipping and returns work."