UCP adoption is a platform effect: what 500 merchants show
Published . Figures computed from the readiness barometer measured on 2026-09-09.
Short answer: 64 of the 500 merchants scanned (9% in France, 16% in the United States) publish a valid Universal Commerce Protocol profile at /.well-known/ucp. 58 of those 64 storefronts carry a Shopify marker, and 59 declare version 2026-08-25, the version Shopify’s changelog says its storefronts advertise by default. Eight months after the protocol launch, UCP discovery on the open web is almost entirely a decision taken by one platform on behalf of its merchants, not a wave of merchant implementations.
What was measured
The barometer requests /.well-known/ucp on the resolved origin of each merchant, with an identified user agent, and validates the body: it must be JSON with a ucp member that carries version and services, as the specification requires for a business profile. A bare HTTP 200 is not enough; 24 sites in the panel return 200 with a page that is not a profile, typically their homepage served under any path. For each valid profile, a second request checks the storefront headers and the profile text for a Shopify marker. The panel, the exclusion list and the scanner are published with the barometer.
The numbers
- France: 23 valid profiles out of 250 merchants (9%), 22 with a Shopify marker. Versions: 2026-08-25 (22), 2026-01-23 (1).
- United States: 41 valid profiles out of 250 merchants (16%), 36 with a Shopify marker. Versions: 2026-08-25 (37), 2026-01-23 (3), 2026-04-08 (1).
- Profiles without a Shopify marker, all countries: 6. They are arhaus.com (2026-08-25), academy.com (2026-01-23), chewy.com (2026-01-23), scheels.com (2026-04-08), sezane.com (2026-01-23), backcountry.com (2026-01-23).
- Category with the highest share of valid profiles: beauty and cosmetics in France (23%), beauty and cosmetics in the United States (51%).
Why the version tells the story
UCP profiles are dated by release. The specification’s business profile example shows ucp.version alongside the services and capabilities the business supports. On 25 August 2026 the project shipped release v2026-08-25, with multi-vertical structures, 3DS2 flows and breaking schema changes. Shopify’s developer changelog announced that “Shopify storefronts now advertise support for Universal Commerce Protocol (UCP) version 2026-08-25 in their discovery profiles at /.well-known/ucp”, adding that the update “doesn’t require any changes to your existing integrations” and that “the set of capabilities that Shopify supports is unchanged”. A merchant on Shopify therefore woke up one morning with a fresh profile, without having decided anything.
That is exactly what the panel shows. The version distribution is not a spread of merchant decisions taken at different times; it is one platform’s release calendar reflected on 59 storefronts. The few profiles that carry other versions (arhaus.com (2026-08-25), academy.com (2026-01-23), chewy.com (2026-01-23), scheels.com (2026-04-08), sezane.com (2026-01-23), backcountry.com (2026-01-23)) are the ones worth studying if you want to see what a merchant-side or non-Shopify implementation looks like in the wild, and there are very few of them in a panel of the 500 most visited commerce sites of two countries.
What a profile does and does not prove
A discovery profile is an advertisement of protocol version, services, capabilities and payment handlers. It is the first step of negotiation between an agent platform and a business. It does not prove that an agent can complete a purchase at that merchant today: the agent platform must support the same version, the capabilities must be enabled, and the surfaces where agent checkout is offered (Google’s AI Mode and Gemini were named at launch) must have opened to that merchant. None of that is visible from the outside, which is why the barometer labels the tier “agent-ready signals”, not “agent-ready”.
The reverse is also true. A large retailer that runs its own commerce stack can be integrated with an agent platform through a private channel and show nothing at /.well-known/ucp. Three of the panel’s US retailers were named as UCP co-developers in Google’s launch announcement, Walmart, Target and Wayfair, and none of them serves a profile at its public origin (target.com: /.well-known/ucp HTTP 404, walmart.com: /.well-known/ucp HTTP 404, wayfair.com: /.well-known/ucp HTTP 404). Absence of a public signal is not absence of work.
What this means for a merchant
If you are on Shopify, you already publish a profile and the question is not “should I adopt UCP” but “which capabilities are actually enabled for my store, and on which agent surfaces”. Check the profile your storefront serves, read Shopify’s agent-profile documentation, and test the negotiation from a sandbox agent before you tell anyone your store is agent-ready.
If you are not on Shopify, the barometer says you are in the large majority, and that the reference implementations you can look at are rare. Start with the specification’s business profile, publish only the capabilities you can honour, keep your product, price, availability and policy data accurate first (an agent that reads a profile will read those next), and run the readiness diagnostic before the profile. The implementation guides list the documented surfaces per platform.
Limits
One pass, one date, public signals only. A profile served on a checkout subdomain or behind a bot wall is not seen. The Shopify attribution relies on response headers and on the profile text, which can be masked by a proxy; the count is therefore a floor, not a ceiling. The next issue of the barometer, October 2026, will replay the same 500 addresses and report the change.
Questions
How many merchants publish a UCP profile in 2026?
/.well-known/ucp, meaning a JSON document with ucp.version and ucp.services. HTTP 200 responses that were not profiles (24 sites) are not counted.