Specoria
Menu
TR Free test

InsightsAgentic commerce

97 of 100 big US retailers have no checkout for AI agents. Should your store worry?

A new vendor study says almost no large US retailer offers AI agents a way to buy. Why that is less alarming than it sounds, and what matters more right now.

By 5 min read
A cut-paper collage with an empty paper browser frame in the center, a plaster hand reaching toward a row of grey shopping bags, one bag in violet, and a torn stone arch behind them.
A shop nobody names is a door no one walks through. Illustration: Specoria · generated with AI

On 6 October 2026, the commerce software company Rezolve Ai published its US AI Visibility Study 2026. It looked at how four AI assistants present 2,047 US companies, and one line travelled fast: 97 of America's 100 largest retailers publish no way for an AI agent to check out.

It is a striking number. Before treating it as a deadline for your own store, it helps to look at what the study measured, what has happened to agent checkout this year, and where agents actually lose sales today.

What the study says, and what it does not

The finding comes from a vendor that sells tools for this exact problem, so it is the vendor's own measurement. The press release, distributed through GlobeNewswire, does not say how a "way to check out" was detected. It points to a separate methodology page that we have not reviewed. Until that method is public, the safest reading is narrow: large retailers have not published a dedicated, machine-readable checkout route for agents.

The same release reports a second result that matters just as much. One in four of the 2,047 companies is never named by the assistants it tested (OpenAI, Claude, Perplexity and Gemini). A store that is never recommended does not get to the checkout question at all.

Agent checkout had a rough year

A cut-paper collage with a plaster hand holding a fan of partly covered price tags above a winding path of paper stepping stones, leading to a small stone arch and a closed parcel, with one violet paper key on the path.
Before the basket, the road must be clear and every tag must tell the same story. Illustration: Specoria · generated with AI

The 97 figure lands in a year when the most visible form of agent checkout went backwards.

Walmart offered roughly 200,000 products through OpenAI's Instant Checkout, which let people buy inside ChatGPT without visiting Walmart's site. In March 2026 a Walmart executive told WIRED that purchases made inside ChatGPT converted at about one third of the rate of shoppers who clicked through to Walmart's website, and that the company was moving away from the in-chat model. OpenAI confirmed it was phasing Instant Checkout out in favour of checkout handled by the merchant inside its own app.

So the large retailers in the study are not simply asleep. Many of them tried the first version, saw weak results, and are waiting for a model where the merchant keeps the customer relationship and the payment step.

Where agents lose sales today

For most stores, the purchase is not lost at a missing checkout API. It is lost earlier and more quietly:

  • The agent never finds the store. Assistants recommend what they can read and trust. If product pages block AI crawlers in robots.txt, render prices only with JavaScript, or carry no structured product data, the store drops out before any shopper sees it.
  • The agent cannot confirm the details. Price, stock, variants, delivery time and return window are what a shopper asks about. When the page, the structured data and the product feed disagree, an agent hesitates or picks a competitor whose data agrees.
  • The browser agent gets stuck on the way to pay. Agents that use a real browser, the most common kind today, still have to get through forced account creation, cookie walls, address forms and bot challenges. Each one is a place where the task can end.

None of these needs a new protocol. They are fixable on the store's own pages, and they also decide whether a future agent checkout would have anything to build on.

Should you build an agent checkout now?

For most stores, not yet. The protocols are still settling: in-chat checkout in ChatGPT was rolled back, Google's Universal Commerce Protocol is live in the US, and identity standards such as the Personal Agent Protocol announced on 6 October are only at the specification stage. Building a dedicated agent checkout today means guessing which of these will matter for your customers.

What pays off now is the groundwork every one of these approaches will need:

  1. Let shopping agents in. Allow the search and user agents of the major assistants in robots.txt, and check that your CDN or firewall does not block them silently.
  2. Make product data agree everywhere. The same price, stock status and variants on the page, in the structured data and in any feed you send to marketplaces or assistants.
  3. Publish policies a machine can read. Return window, shipping cost and delivery time as structured data, not only as a paragraph on a help page.
  4. Keep the human checkout reachable. Guest checkout, plain form fields and no extra hurdles before payment help browser agents and people alike.
  5. Measure what agents see. A store that knows where an agent stops can fix the right thing first.

The bottom line

"97 of 100 have no agent checkout" is a useful wake-up call, but the more important line in the same study is the one about visibility: a quarter of companies are never named at all. Before a store worries about how an agent pays, it should make sure an agent can find it, read it and get through its existing checkout.

Specoria measures exactly those steps: whether agents can reach your pages, whether your product and policy data agree, and where a real browser agent stops on the way to the basket. Start with a free readiness test.

Sources