Engineering
June 30, 2026

Who owns the Checkout? Merchant-controlled Agentic Commerce

Who owns the Checkout? Merchant-controlled Agentic Commerce

The 2026 evidence is in - and clearer than anyone expected. Here's why merchant-controlled architecture wins, what the new protocols actually do, and what European merchants should build.

Merchant-controlled checkout won agentic commerce in 2026. ChatGPT Instant Checkout converted approximately 3× worse than direct merchant checkout. Walmart Sparky lifted average order value 35%. The architecture that's emerging: merchants own the checkout endpoint and route protocol mechanics (ACP, UCP, AP2, x402) through a facilitator.

Two visions of "checkout" competed in the early years of agentic commerce. In one camp: the agent owns the transaction end-to-end, sweeping the user through to "purchase complete" without ever leaving the chat. ChatGPT Instant Checkout launched on that thesis. In the other camp: the agent helps with discovery and intent, but the merchant owns the checkout surface and the customer relationship. Walmart Sparky, Shopify's agent flows, and the production deployments at Tatcha and others took that path.

By 2026, the data settled the debate. The merchant won.

What does the 2026 data actually show?

ChatGPT Instant Checkout converted approximately 3× worse than direct merchant checkout on the same product searches (Walmart says ChatGPT Checkout converted 3× worse than its own website, MarTech, 2026). By early 2026, OpenAI was publicly retreating from the in-chat checkout model, pivoting Walmart and other partners back toward "agent recommends, merchant fulfils."

On the other side of the ledger, Walmart Sparky - Walmart's own AI assistant operating inside their checkout drove a 35% lift in average order value (Walmart leadership commentary, Q4 2025 earnings call). Tatcha, the skincare brand, reported 3× conversion improvement after integrating agentic features into their own site rather than ceding them to third-party agents (Own case study).

Three data points, three different verticals, same direction. The market had answered the question.

Why do merchants own the checkout?

The conversion-rate story is the headline, but the deeper reasons are structural:

Inventory and pricing accuracy. When an agent shows the user a €50 product and tries to check out, the merchant may have updated the price, run out of stock, or shifted shipping zones in the seconds since the agent crawled. Only the merchant can render the canonical state at the moment of transaction.

Brand and experience. Merchants spend years building trust around a checkout experience – loyalty programs, return policies, refund flow, customer-support handoff. Surrendering that to an agent surrenders the loyalty relationship.

Compliance and regulatory cleanliness. Particularly relevant for European merchants: VAT calculation, GDPR data handling, MiCA-compliant settlement for stablecoin transactions, and consumer-protection law around digital goods all sit with the merchant of record. Agents can't take that on without re-architecting their compliance stack.

Credential safety. When the agent takes a credit card, the agent becomes a card-data handler, a security and compliance nightmare. Merchant-controlled checkout means the merchant's PCI scope stays intact and the agent never sees raw payment credentials.

Each of these alone might be marginal. Together they're decisive which is what the conversion data was always reflecting.

Merchant ownership of checkout doesn't mean merchant ownership of every protocol primitive underneath. The parallel is Stripe in 2012: merchants kept their store, brand, and customer; Stripe owned the boring middle i.e. PCI compliance, network routing, retry logic, dispute handling, for 2.9% + 30¢. The agentic-commerce stack needs the same layer. ACP, UCP, AP2, and Shared Payment Tokens are protocol surfaces, not turnkey implementations. A facilitator absorbs production-grade x402 mechanics, KYA logging, MiCA-compliant settlement, and refund handling so the merchant keeps the checkout without owning the protocol plumbing.

What protocols sit underneath merchant-controlled checkout?

If merchants own the checkout, the protocols agents speak need to negotiate with merchants rather than impersonate them. Four primitives sit at this layer.

ACP: Agentic Commerce Protocol

ACP is a checkout handshake protocol developed by Stripe and OpenAI.

The agent presents an intent, the merchant returns the actual cart and price, the user (or an authorised agent mandate) confirms, and settlement happens through the merchant's existing rails. The agent never sees the cart line items or the card number. ACP has been live with Etsy since February 2026, with 1M+ Shopify merchants (including Glossier and SKIMS) rolling out through 2026.

UCP: Universal Commerce Protocol

UCP is a full-lifecycle agent-to-merchant protocol developed by Google, Shopify, and the Universal Commerce Protocol Council.

The Council, founded in April 2026, includes Amazon, Meta, Microsoft, Salesforce, and Stripe. UCP covers product discovery, capability negotiation, cart building, and order tracking, the full agent-to-merchant lifecycle of which checkout is one stage. ACP and UCP are complementary rather than competitive: UCP describes the merchant's full API surface; ACP handles the specific checkout step within that surface.

AP2: Agent Payments Protocol

AP2 is a mandate-governance protocol for agentic payments, originally developed by Google and donated to the FIDO Alliance in Q2 2026 for community governance.

It defines who is authorised to check out, with what limits, on whose behalf, so the merchant has cryptographic proof of authorisation before accepting payment. A Mastercard-co-developed "Verifiable Intent" standard layers into the mandate spec. AP2 wraps the mandate layer beneath both ACP and x402, with 60+ founding organisations including Mastercard, Adyen, PayPal, and Revolut.

Shared Payment Tokens

A Shared Payment Token is a one-time, scoped credential the merchant issues at checkout that the agent passes through without holding the underlying card or wallet details.

Mastercard and Visa both shipped versions in 2026 (Mastercard Agent Pay and Visa Intelligent Commerce). x402's on-chain equivalent is the signed mandate plus settlement-on-execution model: agent has authorisation, merchant has proof, no one has raw secrets in flight.

What this means for European merchants?

ACP, UCP, and AP2 are all US-led standards, and they implicitly assume USD settlement on card rails. European merchants face additional structural complexity: MiCA-compliant stablecoin settlement, SEPA-native fiat flows, GDPR data handling, and (where regulated) PSD2 authentication.

The right architecture for an EU merchant adopting agentic commerce in 2026:

The merchant owns the checkout endpoint, implementing UCP-compatible product schema and ACP-compatible checkout handshake. The merchant uses Shared Payment Tokens (Visa Intelligent Commerce, Mastercard Agent Pay, or x402's signed mandates) so the agent never sees raw credentials. Settlement happens in a MiCA-compliant stablecoin for stablecoin-native flows: USDC and EURC via Circle, or EURAU via AllUnity for EU-anchored merchants who want BaFin-EMI-issued settlement. An x402 facilitator handles the production-grade error handling, retries, receipts, and KYA logging that the bare protocols leave to the implementer (see "x402 in production: what the protocol leaves out" for the five gaps every implementation hits).

Bluerails as the European facilitator

Bluerails is the European x402 facilitator stack, AllUnity-partnered for EU EUR settlement.

Bluerails handles the protocol handshake (ACP / UCP / AP2), the production-grade error layer (retries, receipts, KYA logging), and the MiCA-compliant settlement leg via AllUnity. Bluerails never holds your customer relationship, your cart, or your brand. You issue the Shared Payment Tokens. You confirm the cart. You own the receipt. Your checkout. Our infrastructure underneath it.

The merchant of record stays the merchant. The agent does the discovery and the authorisation. Nobody loses their checkout to anyone.

How do you implement this in 6 steps?

  1. Build the agent-payable endpoint at your domain - merchant.com/agent-checkout or equivalent.
  2. Implement UCP-compatible product and inventory schema so agents can discover and price your offerings reliably.
  3. Wire ACP for the checkout handshake - intent, cart resolution, confirmation, settlement.
  4. Use Shared Payment Tokens or signed mandates, never expose raw credentials to the agent.
  5. Wrap the auth flow with AP2 mandates so you have cryptographic proof of who authorised what.
  6. Settle in a MiCA-compliant stablecoin via a production x402 facilitator, for EU flows, EURAU via AllUnity is the cleanest EUR-native option.

Steps 1–3 are merchant-side engineering, 1–2 weeks if you start from scratch. Steps 4–6 collapse to under an hour if you route through a facilitator that has already plumbed the standards, which is what Bluerails ships.

Frequently asked questions

Who owns the checkout in agentic commerce?

The merchant owns the checkout. Production data through 2026, including a 3× conversion gap between ChatGPT Instant Checkout and direct merchant checkout has settled the question. Agents handle discovery and authorisation; merchants own the transaction surface.

What is ACP?

ACP (Agentic Commerce Protocol) is a checkout handshake protocol developed by Stripe and OpenAI. It powers the "Buy it in ChatGPT" / Instant Checkout flow live with Etsy and rolling out to 1M+ Shopify merchants.

What is the difference between ACP, UCP, and AP2?

ACP handles the checkout-step handshake between agent and merchant. UCP covers the full agent-to-merchant lifecycle (discovery, cart, order tracking). AP2 is the mandate-governance layer that wraps both. They are complementary, not competing.

What is a Shared Payment Token?

A Shared Payment Token is a one-time, scoped credential a merchant issues at checkout that an agent passes through without holding the underlying card or wallet details. Mastercard Agent Pay and Visa Intelligent Commerce are the two card-network implementations; x402 has an on-chain equivalent via signed mandates.

How do European merchants accept agent payments under MiCA?

European merchants own the checkout endpoint, use Shared Payment Tokens or x402 signed mandates so the agent never sees raw credentials, and settle in MiCA-compliant stablecoins (USDC, EURC via Circle, or EURAU via AllUnity). An x402 facilitator like Bluerails handles the production protocol mechanics and MiCA-compliant settlement so the merchant doesn't need to become a CASP.

Why did ChatGPT Instant Checkout retreat?

ChatGPT Instant Checkout retreated in early 2026 because it converted approximately 3× worse than direct merchant checkout on the same product searches. OpenAI publicly pivoted partners back toward "agent recommends, merchant fulfils" architecture by early 2026.

What's next

The interesting battleground in 2026–2027 isn't who owns the checkout, that's settled. It's where the merchant publishes the agent-payable endpoint so agents can find it in the first place. That's the discovery layer (see "How AI agents actually find and trust your site"), and where vertical registries like Bluerails' hotel and publisher registry start to matter.

If you want to see where your site stands today, Bluerails' Agent Score scanner runs the checklist above and tells you exactly what's missing. Drop in a URL, get the score, get the UCP + ACP + x402 endpoint code generated for you that ships in under an hour with AllUnity settlement plumbed underneath.

Heading 1

Heading 2

Heading 3

Heading 4

Heading 5
Heading 6

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.

Block quote

Ordered list

  1. Item 1
  2. Item 2
  3. Item 3

Unordered list

  • Item A
  • Item B
  • Item C

Text link

Bold text

Emphasis

Superscript

Subscript