Declared is not payable.

Supply is arriving. Payability is not.

Shopify adopted the new spec in six days. Wix is still on the old one.

Two platforms, opposite strategies

The cliff is between announcing and transacting

What happened when we tried to buy something

• Shopify blocked our own store from autonomous checkout

  probed 25 June 2026
  We built an agent client, authenticated at Shopify’s highest self-serve tier, and ran live checkout probes against third-party merchants and against our own store. Every path returned checkout_completion_ineligible, including our own. The own-store control is the point: it separates "not rolled out yet" from "gated to Participating Channels".

• The first native Wix UCP storefronts shipped empty payment handlers

  probed 29 June 2026
  Within days of the first native Wix manifests appearing we probed the cohort store by store: 10 storefronts across 9 distinct sites. Every one declared the same four capabilities (checkout, fulfillment, catalog search, catalog lookup) and every one had an empty payment_handlers array. Several returned unauthorized; others were unconfigured, one shipping a sample product literally titled "I’m a product".

• Two platforms hold 99% of agent-ready supply

  probed 2 September 2026
  A June fleet scan across 3,000 candidate domains skipped 2,652 as Shopify, and we reported roughly 88% on one platform with a thin tail. That was wrong, and the reason is worth stating: the scan seeded from a published export whose rows were last checked between 28 March and 2 April, months before either current spec existed. Read directly on 2 September, the split is Shopify 10,464 and Wix 5,576 of 16,175 verified. Non-Shopify supply is about a third of the market, and almost all of it is one company.

• A third of agent-ready supply declares no way to pay

  probed 2 September 2026

  We fetched manifests for the 25 highest-scoring Wix storefronts and the 21 highest-scoring Shopify ones. Neither set varied internally at all.

  • Wix, 25 of 25. Spec 2026-04-08, the same four capabilities every time, and payment_handlers: []. The key is present and the array is empty.
  • Shopify, 21 of 21. Spec 2026-08-25, eight capabilities, and populated handlers on every store (dev.shopify.card, com.google.pay, dev.shopify.shop_pay).

  No Wix store declared supported_versions, so none of them can negotiate to 2026-08-25. Results this uniform are a platform default. Thousands of merchants did not each arrive at the same configuration on their own.

  One bound on this: we sampled by UCP Score, so a payable Wix storefront would probably have outranked the ones we read. The reading supports a claim about Wix defaults and about the top 25. It does not establish the figure for all 5,576.

• The distributor side is walled at settlement

  probed 30 June 2026
  CJ Dropshipping settles only from a CJ-held balance with no API top-up path, so every payment mode is platform-fronted. Mouser requires an account opened on the buyer’s behalf. Both are reachable by an agent and neither is payable by a neutral one.

• Declared AP2 support is mostly developer test sites

  probed 11 July 2026
  Of the handful of domains declaring AP2, most are developer or sandbox targets, and none properly enabled checkout against it when probed. Declared is not payable, and the gap between the two is the entire finding.

What is wrong with this data

• The grade distribution does not discriminate. In the 2026-08-31 reading, 14,690 of 15,942 verified stores score an A, 1,251 score a B, exactly one scores a C, and nothing scores below that. A score where 92% of the population lands in the top band tells you a manifest exists and little else.
• Capability counts exceed the verified merchant count, so the two are drawn from different populations. We compare tiers against each other and never present them as a funnel.
• The series is transcribed by hand from a weekly public report, and two readings contain upstream corrections we have left in place rather than smoothed.
• Conduit probe findings on this page were run in two waves: 2026-06-25 to 2026-07-11, before the 2026-08-25 spec landed, and 2026-09-02, after it. Each carries its own date and the older wave should be read as a snapshot of the earlier spec era.
• Platform attribution comes from the source. It assigns each verified store to a platform from manifest signals and HTTP headers, and we have not independently verified that mapping. The spec version and payment-handler readings in the September findings are ours, taken from the manifests directly.
• The September manifest reads sampled 25 Wix and 21 Shopify storefronts ranked by the source’s own score. A high-ranking sample is the right place to look for a capability that exists, and the wrong place to estimate how common one is.