Loading...
Loading...
A quarterly snapshot of agentic commerce. UCP Tech Council expands. ACP scales back checkout. Microsoft launches Copilot Checkout. What is working, what is not, and what to watch next.
Agentic commerce moves on a faster cycle than most retail trends, so a quarterly check-in is less of a luxury than it used to be. In the eight weeks since our last industry-level write-up, the UCP Tech Council expanded by five major members, the protocol shipped its largest single release, OpenAI publicly scaled back its in-chat checkout ambitions, Microsoft turned Copilot into a buy lane, and Walmart pulled its own AI assistant inside ChatGPT.
This piece is the snapshot. What actually happened between February and late April 2026, what is working, what is not, and what to watch through the summer.
Honest framing first: the public adoption data for agentic commerce in early 2026 is uneven. We have a few solid forecasts and post-holiday measurements; we do not have rigorous quarterly transaction telemetry yet. Three numbers are worth anchoring on.
eMarketer projects AI platforms will account for roughly 1.5% of total US retail ecommerce sales in 2026 — about $20.9 billion, nearly quadruple 2025 levels. The same forecast pencils in $144 billion (8.8% of total) by 2029. That is the "real but small" curve.
Salesforce reported that AI and agents drove approximately 20% of retail sales during the 2025 holiday season. That figure mixes everything from "shopper used AI to research before buying" to "agent placed the order." The headline is high; the share that is fully autonomous is much smaller.
Adobe reported generative AI referral traffic surged 693% during the 2025 holiday season — the clearest signal that AI surfaces are becoming a meaningful traffic source even where they are not yet meaningful transaction sources.
The honest read across all three: agentic commerce is past the demo phase, the discovery and influence loops are working, and end-to-end autonomous purchase volume is still small enough that 1-2 large merchants moving meaningfully changes the trajectory.
The single biggest event of the quarter was the April 24 expansion of the UCP Tech Council. Amazon, Meta, Microsoft, Salesforce, and Stripe joined the existing five (Google, Shopify, Etsy, Target, Wayfair). Three of those five names — Microsoft, Stripe, Salesforce — were until recently aligned with OpenAI's ACP. One — Amazon — has its own well-developed agentic shopping stack (Rufus reached 300 million users and drove $12 billion in incremental sales in 2025, per Newsfile reporting on the announcement). One — Meta — controls a massive social commerce surface.
This is the move that converts UCP from "Google and Shopify's protocol" into "the place where commerce protocol disagreements get resolved."
Vidhya Srinivasan, Google's VP/GM of Ads and Commerce, framed it carefully: "The Universal Commerce Protocol is quickly paving the way for this new era of agentic commerce. We're proud to see the industry come together around this shared, open standard that will benefit businesses and consumers everywhere." Vanessa Lee, Shopify's VP of Product, was similarly measured: "AI can enable so many new ways of shopping to flourish, but only if there's a clear standard between retailers, businesses and applications."
The governance question is the right one to track. The Tech Council "aligns UCP's technical direction, reviewing contribution proposals and stewarding the open-source protocol." Whether that produces durable consensus when, for example, Amazon proposes capabilities that look more like Amazon's stack and less like Google's stack is the open question for the rest of 2026.
For a technical breakdown of the April spec release that immediately preceded the Tech Council expansion, see our UCP v2026-04-08 update post.
Two weeks before the Tech Council expansion, UCP shipped v2026-04-08 — the largest single release in the protocol's history. The spec added Cart, Catalog, and Identity Linking capabilities; introduced request and response signing; tightened error handling on capability negotiation; and made currency a required field on Order responses.
This matters because the original v2026-01-23 spec, what most early UCP merchants implemented at NRF, was effectively a single-item guest checkout. You could buy "this Etsy item" through Google AI Mode, but you could not assemble a multi-item cart, look up live variants, or get personalized loyalty pricing. The April release fixes all three. Cart enables baskets. Catalog enables real-time product queries. Identity Linking enables logged-in personalization.
Shopify followed up on April 22 by shipping the Storefront Catalog MCP that implements the new UCP Catalog capability, with the previous tool versions kept alive until June 15, 2026. Google updated Merchant Center onboarding the same week. Salesforce, Stripe, and Commerce Inc were named as platforms implementing the new capabilities.
Net effect: by mid-April, UCP went from "interesting standard with one live use case" to "interesting standard with broad platform implementations and the building blocks of a real storefront."
The most under-discussed event of the quarter was OpenAI's March 24 walk-back of Instant Checkout. The official framing was diplomatic — "the initial version of Instant Checkout did not offer the level of flexibility that we aspire to provide." TechCrunch's reporting was less diplomatic: ChatGPT users were not actually buying through Instant Checkout in meaningful volume, and partner merchants reported minimal sales from ChatGPT referrals.
OpenAI's response was to refocus ACP on product discovery. The new ACP gives merchants a structured feed format for their catalogs and renders products visually inside ChatGPT, with side-by-side comparison, conversational filtering, and image-based lookalike search. Checkout still exists, but the user is routed to the merchant's site (in-app browser on mobile, separate tab on desktop). It is a shopping research tool more than a shopping transaction tool.
Major retailers integrated for ACP discovery as of late March 2026 include Target, Sephora, Nordstrom, Lowe's, Best Buy, The Home Depot, and Wayfair. Walmart's "Sparky" experience inside ChatGPT, announced March 25, runs on the same model.
This is healthy. ACP-as-discovery is a far better fit for OpenAI's actual conversion data than ACP-as-checkout was. It also means the protocol space is consolidating: ACP no longer competes with UCP on the checkout axis, only on the discovery axis. And on the discovery axis, ACP's structured feed model (snapshot product feeds, daily refresh, promotions API) overlaps directly with UCP's new Catalog capability.
For a deeper look at why UCP and ACP make different bets on the same ground, see our UCP vs ACP comparison and the four-protocol landscape explainer.
Microsoft launched Copilot Checkout on January 8, 2026 with PayPal, Shopify, and Stripe as core partners. Urban Outfitters, Anthropologie, Ashley Furniture, and Etsy were the named consumer brands at launch. Shopify merchants were auto-enrolled with an opt-out window — a quiet mass enrollment of millions of stores into Microsoft's checkout flow.
Microsoft cited internal data showing shopping journeys involving Copilot are 33% shorter than traditional search paths and produce a 53% increase in purchases within 30 minutes of interaction. Those numbers are Microsoft's, on Microsoft's funnel, and they should be read with appropriate skepticism — but the direction is consistent with the Salesforce holiday data and the Adobe traffic data.
What is interesting is what Copilot Checkout did not adopt. It is not built on UCP. It is not built on ACP. It is its own commerce flow with PayPal and Stripe rails. Microsoft's April 24 entry into the UCP Tech Council suggests that may change — or at least that Microsoft now wants influence over how UCP evolves to interop with Copilot — but as of the end of April, Copilot Checkout remains a parallel third lane.
For merchants, the implication is straightforward. If you sell on Shopify, your store is already in Copilot Checkout (unless you opted out), already in Storefront MCP, and almost certainly already serving its UCP manifest at /.well-known/ucp. Three rails, one platform, no extra integration work.
Shopify is the merchant in this whole story most quietly winning. The company sits inside the UCP Tech Council, has its catalogs surfaced through Storefront MCP for any MCP-capable agent (Claude, ChatGPT, Gemini, Copilot), is auto-enrolling stores into Copilot Checkout, and is the headline platform partner for Google's UCP rollout in Google AI Mode and Gemini.
The Shopify position is best summarized as: every Shopify merchant gets agentic commerce by default, on every major surface, without doing anything. The Storefront Catalog MCP update on April 22 added UCP-native search_catalog, lookup_catalog, and get_product tools, with the old endpoints maintained until June. The Hydrogen Winter 2026 Edition pulled Storefront MCP, Catalog, and Dev MCP into headless storefronts on Oxygen.
Shopify's Q1 2026 numbers are not yet public at the time of writing. The earlier February analysis from Shopify reported 42% merchant adoption of its AI features (Sidekick, Magic, etc.). Whether that translates into meaningful incremental GMV through agentic surfaces is the headline number to watch in Q2.
The single most validated agentic commerce surface today is Google AI Mode and the Gemini app, where UCP-powered checkout has been live for Etsy and Wayfair items since February 11. Sundar Pichai previewed this at NRF in January with a "soon you'll see a buy button directly on Google surfaces" line; that buy button is now operational.
Google has confirmed Shopify merchants, Target, and Walmart as the next checkout integrations. Hundreds of additional companies are in active integration discussions according to Google's February announcement. The April 8 spec update with Cart and Identity Linking is what makes those expansions feasible — the original spec literally could not handle them.
The other Google development worth tracking is Direct Offers and the new Sponsored ad format inside AI Mode. Both turn UCP-mediated discovery into a paid surface, which is the obvious commercial future. Google Marketing Live, Google's annual ads event, will be the next inflection point for those announcements.
Five things are working in agentic commerce right now.
Discovery referral. Adobe's 693% generative AI referral traffic growth is the cleanest signal. Shoppers are starting their journeys in AI surfaces in volume.
Cross-platform availability. Shopify's universal Storefront MCP rollout, plus auto-enrollment in Copilot Checkout, plus auto-presence in Google AI Mode through UCP, means a typical Shopify merchant is now reachable through every major AI surface without having had to integrate with any of them.
Merchant-of-record clarity. Every protocol — UCP, ACP, Copilot Checkout — has converged on the merchant-of-record model. Stores keep customer data, fulfillment, and customer relationships. AI surfaces handle the discovery and the rails. This is the single biggest factor that has unblocked enterprise retailer participation.
Spec governance. UCP's Tech Council expansion is not a dramatic move, but it is the move that makes UCP a credible long-term standard. Until April, the realistic worst case was "Google walks away and the protocol orphans." That worst case is now meaningfully harder to imagine.
Infrastructure honesty. OpenAI publicly scaling back Instant Checkout was awkward but useful. It removed the ambient pressure to pretend that in-chat checkout was working better than it was, and gave the rest of the industry permission to focus on what is actually converting (discovery, comparison, handoff to the merchant's own checkout).
Three things are honestly not working yet.
End-to-end autonomous purchase at scale. AP2 mandates exist; they are not in widespread production use. The vast majority of UCP transactions today are still human-in-the-loop, with a person reviewing the cart and approving the payment. The promise of "tell your agent to buy you running shoes and it just does it" is technically possible and consumer-facing implementations remain limited.
Full-funnel measurement. Merchants do not yet have a clean way to see how their products are performing inside agent surfaces. Catalog-level dashboards, conversational impression metrics, and AI-surface attribution are all early. eMarketer's January coverage explicitly flagged retailer pushback on this issue as a 2026 risk to AI-shopping growth.
Protocol convergence with OpenAI. UCP and ACP are now overlapping on discovery without being aligned on it. OpenAI is not on the UCP Tech Council. Stripe is on both sides. Whether OpenAI eventually joins UCP, builds a bridge between ACP feeds and UCP Catalog, or continues to operate a parallel ecosystem is the most important open question for merchants who care about reach.
Five concrete things to watch over the next eight weeks.
Agentic commerce in April 2026 is real, growing, and considerably less hype-shaped than it was at NRF in January. The forecast numbers are credible if modest. The protocol picture is dramatically clearer than it was three months ago. The merchant cost to participate is, for the first time, low for the median merchant (because Shopify, Salesforce, and the platform layer are doing the work) and tractable for everyone else.
What has not happened is the consumer-facing breakthrough that would make agentic commerce feel inevitable to a non-industry observer. ChatGPT did not become Amazon. Google AI Mode is not yet most people's default shopping start. Copilot Checkout is brand-new. The infrastructure is being built faster than the consumer behavior is shifting.
Three months from now, the question worth asking is whether the consumer behavior has caught up to the infrastructure. If it has, 2026 will look like the year agentic commerce went mainstream. If it has not, 2026 will look like the year the rails got built — which is still a meaningful year, just a quieter one than the NRF press cycle promised.
For the merchant directory of stores already shipping UCP, the category browser view by vertical, or a free agentic commerce audit of your own store, those tools are live now and updated continuously as the protocol picture evolves.
According to a December 2025 EMARKETER forecast, AI platforms will account for roughly 1.5% of total US retail ecommerce sales in 2026 — about $20.9 billion, nearly four times 2025 levels. Salesforce reported AI and agents drove around 20% of retail sales during the 2025 holiday season.
Yes. UCP-powered checkout has been live in Google AI Mode and the Gemini app since February 11, 2026 for Etsy and Wayfair items, with Shopify, Target, and Walmart confirmed as next.
On March 24, 2026, OpenAI scaled back Instant Checkout. ChatGPT shopping pivoted toward product discovery, with major retailers including Target, Sephora, Nordstrom, Lowe's, Best Buy, The Home Depot, and Wayfair integrating into the refocused ACP.
As of April 2026, yes. Microsoft launched Copilot Checkout on January 8, 2026 with Shopify, PayPal, and Stripe as launch partners. Microsoft joined the UCP Tech Council on April 24, 2026.
Not yet, but the surface area has narrowed sharply. With Microsoft, Stripe, Amazon, Meta, and Salesforce now on the UCP Tech Council, UCP is the only commerce protocol with credible cross-platform governance.
Three things: get on the UCP v2026-04-08 schema (Shopify and Salesforce do this automatically); supply ACP product feeds if you sell categories where ChatGPT discovery is meaningful; and watch Copilot Checkout as it rolls out — Shopify merchants are auto-enrolled.
Discover the merchants powering autonomous commerce with the Universal Commerce Protocol.