Protocol land grab runs ahead of merchant adoption in agentic commerce

The plumbing is built; the question is who is using it

Eight months into 2026, the plumbing of agentic commerce is largely built. What is less clear is how many merchants are using it, and how much of the traffic flowing through AI surfaces converts into orders that anyone can attribute.

The competitive picture set itself in January. Google announced its Universal Commerce Protocol at NRF, co-developed with Shopify and endorsed by a broad coalition spanning Walmart, Target, Etsy, Best Buy, PayPal, Visa, Stripe and American Express. UCP is scoped across the full journey — discovery, cart, checkout, order management — rather than the checkout call alone, and composes with AP2 for payment authorisation and MCP for tool access.

OpenAI's Agentic Commerce Protocol, built with Stripe and released under an open-source licence, took the narrower route: a checkout layer, now also used by Microsoft Copilot and Shopify's agentic plan. Its core primitive is the shared payment token — a merchant-scoped, amount-scoped, short-expiry token issued by the buyer's payment provider, so the agent never touches saved card credentials.

Most enterprise retailers appear to have declined to choose between the two protocols. Checkout.com's analysis, widely cited since, found that merchants running both protocols captured around 40% more agentic traffic than single-protocol adopters.

The adoption gap

The counterweight to the protocol announcements is merchant behaviour. OpenAI launched Instant Checkout in late September 2025 with Etsy and Shopify merchants as first partners. Forrester principal analyst Emily Pfeiffer put the number of Shopify merchants actually using it at roughly 30 as of February 2026 — against a pipeline routinely described as over a million.

In March 2026 OpenAI retired Instant Checkout in its original form, moving toward dedicated retailer apps inside ChatGPT and, more broadly, toward a model where discovery happens in the assistant and the transaction completes in a merchant-controlled environment. That shift matters for smaller sellers: it lowers the integration bar from "rebuild checkout" to "publish product data an agent can read".

The traffic side is less ambiguous. Shopify reported AI-driven traffic to its stores growing roughly eightfold year on year in the first quarter of 2026, with orders from AI-powered search up close to thirteenfold and average order values about 14% higher than organic search. Adobe measured a 693% year-on-year rise in AI-referred visits to US retail sites over the 2025 holiday season (data 1 November to 31 December 2025), with those visits converting about 31% better than non-AI traffic.

Those multiples come off a very small base. Triple Whale counted 606,489 AI citations across ecommerce queries in the first quarter of 2026, against 164 million referral transactions through traditional channels in the same period. AI is the fastest-growing retail channel and not yet a primary one.

Five rails, one requirement

Five protocol families now shape the space: ACP from OpenAI and Stripe, UCP from Google and Shopify, AP2 for payment authorisation, Visa's Trusted Agent Protocol and Mastercard Agent Pay for agent identity and payments, with MCP — donated to the Linux Foundation's agentic AI effort in December 2025 — as the connectivity layer underneath.

For merchants, the practical consequence is duller than the acronym count suggests. Every one of these rails assumes structured, machine-readable product data: accurate titles, prices, stock, variants, shipping and returns terms, exposed in a form an agent can parse without guessing. Analysts describe rich structured product data as the single largest lever on agentic performance. It has also, conveniently, been a solved problem for a decade.

Attribution remains the open wound. One 2026 analysis found around 70.6% of AI referrals misclassified as direct traffic in standard GA4 configurations, with reported AI traffic undercounted by an estimated three to four times, partly because paid ChatGPT accounts and Gemini's deep research mode pass no referrer data. Merchants can sell through these channels more easily than they can measure them.

Longer-range forecasts from Gartner and McKinsey place the real inflection in agentic spending somewhere between 2028 and 2030. The gap between now and then is the preparation window, and the work in it — clean product data, consistent entity information, AI-readable policies — is measured in weeks rather than quarters.

For the plain-English version of the protocols and what a small business should do about them, see the field guide chapter on getting agent-ready for agentic commerce.

Sources

Google (UCP, NRF January 2026); OpenAI and Stripe (ACP); Checkout.com (dual-protocol traffic analysis, 2026); Forrester (Instant Checkout merchant adoption, February 2026); Shopify (Q1 2026 AI traffic and order data); Adobe Digital Insights (January 2026); Triple Whale (Q1 2026 ecommerce AI citations); First Page Sage (chatbot market share tracker, June 2026); Google I/O 2026; Gartner and McKinsey (agentic spending forecasts). Figures reflect published research available at the time of writing and are dated snapshots of a fast-moving market.