AI shopping agents can now complete the purchase. Most small stores are not ready to be chosen

For most of the history of online retail, the customer was a person with a browser. They scrolled, compared tabs, read reviews and eventually clicked a button. That is no longer the only customer a store has to win. Shopping agents built into chat assistants can now take a request such as “find me a waterproof hiking jacket under $200 that ships by Friday,” narrow the options, and complete the purchase.

The industry calls this agentic commerce, and the plumbing for it is no longer experimental. Checkout inside ChatGPT has been available since September 2025, running on the Agentic Commerce Protocol, an open specification developed by OpenAI and Stripe that covers agent checkout, carts and product feeds, delegated payment and order updates. Several of the largest payment and commerce platforms, including Stripe, PayPal and Shopify, have made agent support available to existing merchants automatically, without a separate application.

Two announcements on October 5 underline how quickly the conversation has moved from whether to how. A release from search vendor Constructor cited Stripe data indicating that 80% of customers using Stripe’s Link wallet had used a chat-based agent in the past month. And ad-tech company InMobi published results from a storefront readiness scanner; by arithmetic on its published figures, about 57% of the storefronts it scored reached its top readiness tier, leaving a large minority that agents would struggle to use.

What an agent actually needs from your store

An agent does not admire photography, respond to urgency banners or scroll a lifestyle carousel. It reads data. It needs to know precisely what a product is, what it costs including shipping, whether it is in stock in the requested size, when it will arrive, and what happens if it is wrong. When any of that is missing or contradictory, the agent does not take a second look. It moves on to a store where the answer is clear.

That is the uncomfortable shift for small retailers. Brand storytelling still matters for human shoppers, and many purchases will keep happening the old way. But in the slice of commerce handled by agents, the winner is often the store whose catalog is the most legible, not the one with the best design.

A readiness checklist for small and mid-sized stores

1. Complete product data

Every product needs a specific title, a full attribute set (size, color, material, dimensions, compatibility), a GTIN or manufacturer part number where one exists, and accurate variant-level inventory. Free-text descriptions are not a substitute for structured fields.

2. Prices that match everywhere

The price in your feed, on your page and at checkout must agree, and shipping costs must be computable before the final step. A mismatch is one of the fastest ways to lose an agent-led sale.

3. Policies written as facts

Return windows, restocking fees, warranty terms and delivery estimates should be stated plainly, in one place, and marked up. “Hassle-free returns” is marketing. “Returns accepted within 30 days of delivery, free return label, refund in 5 business days” is something an agent can use.

4. Structured markup and a clean feed

Product, Offer and MerchantReturnPolicy markup on your pages, plus a well-maintained product feed in your platform or Google Merchant Center, gives agents and search systems the same source of truth. Use Merchant Center diagnostics or your feed tool to flag stale and missing fields on a schedule.

5. Turn on what your platform already offers

If you sell on Shopify or take payments through Stripe or PayPal, check the agentic features in your admin. Support may already be available to you; the work is in confirming what is enabled and testing it.

6. Plan for verification and fraud controls

Agents are a new kind of visitor. Developer documentation for agent-ready storefronts describes agents signing their requests with published keys so merchants can verify who is calling, because a user-agent string proves nothing. Ask your platform how agent orders are authenticated, flagged and reported.

7. Measure agent traffic separately

Agent-initiated orders collapse discovery, comparison and purchase into a single step, which breaks traditional attribution. Tag and report them separately so you can see whether the channel is growing.

What does not change

None of this replaces a good store. Agents still favor merchants with strong reviews, reliable fulfillment and fair prices, and plenty of shoppers will keep browsing for themselves, especially for considered or emotional purchases. Some retailers are experimenting with more immersive browsing for exactly that reason, while making the same catalog machine-readable underneath. The practical goal is to serve both customers at once: a store a person enjoys and a catalog a machine can trust.