How to Verify an Agent-Payment Protocol Claim Before Citing It — the four checks this wiki runs on every page

Every other page in this wiki gets checked against four tests before a claim survives: is there a real spec (not just an announcement), has anyone outside the issuing company shipped with it, does every number carry its base, and is a "launch partner" being confused with a "shipped integration." This page is the procedure — apply it before citing any protocol claim.

Step 1: Find the Primary Spec — Record Version + Date, Not the Blog Post

An announcement blog post and a specification are different artifacts; the spec is what matters. Concretely:

- AP2 (Google): spec published on GitHub at https://github.com/google-agentic-commerce/AP2, version v0.2. The documentation at https://ap2-protocol.org/ confirms "v0.2" and that the spec is hosted on GitHub, with a FIDO donation announced April 28, 2026. AP2 itself was announced September 16, 2025 (https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol) — announcement and spec-version dates are seven months apart, which is itself the point: don't cite the announcement date as the spec date. See Google AP2 — the one agent-payment authorization spec that shipped code, not just a press release. - Agentic Commerce Protocol (ACP, Stripe/OpenAI): spec on GitHub at https://github.com/agentic-commerce-protocol/agentic-commerce-protocol, version date-based (2026-04-17). The README badges mark it "Beta," with active development continuing in an unreleased/ directory — not a stable release. See Agentic Commerce Protocol (ACP) — the checkout-layer spec Stripe and OpenAI ship, and the deployment it already killed. - Visa Intelligent Commerce: API docs at https://developer.visa.com/capabilities/visa-intelligent-commerce state the platform is "in the process of development and deployment" — i.e. no fixed release date, not a version-tagged spec. See Visa Intelligent Commerce: Tokenized Agent Credentials — the spec-vs-announcement test applied to Visa's own numbers. - Mastercard AP4M: announced June 10, 2026 (https://www.mastercard.com/us/en/news-and-trends/press/2026/june/mastercard-launches-agent-pay-for-machines.html), but the full protocol specification is not publicly available — the developer-docs endpoint (https://developer.mastercard.com/mastercard-track-buyer-payment-agent/documentation) exists but the accessible spec is incomplete. See Mastercard AP4M — an announced authorization layer, not yet a published spec or a shipped integration.

Rule of thumb

if you can't name a version tag and a URL that hosts it, you have an announcement, not a spec — say so explicitly rather than treating the two as interchangeable.

Step 2: Check for an Implementation Independent of the Issuing Company

A spec is only credible once someone outside the company that published it has used it in production. Three categories, in ascending order of evidentiary weight: (a) the issuing company's own reference implementation — expected, not proof of real usage; (b) a real merchant, wallet, or third-party processor that has shipped code using the protocol; (c) transaction evidence — a URL showing the system actually working.

- ACP / ChatGPT Instant Checkout clears all three: OpenAI's own announcement states ACP is "immediately live in ChatGPT for U.S. shoppers through Etsy and soon Shopify merchants" (https://openai.com/index/buy-it-in-chatgpt/); Stripe's press release confirms it issued a Shared Payment Token (SPT) primitive for the deployment (https://stripe.com/newsroom/news/stripe-openai-instant-checkout). - AP2: Etsy is named among 60+ launch partners in the September 2025 announcement, but whether Etsy has a shipped AP2 integration versus merely launch-partner status is unresolved in the sources checked — mark this (unverified) rather than asserting a shipped integration. - Coinbase Agentic Wallets launched February 11, 2026 with native x402 support and MPC-secured wallet infrastructure (https://genfinity.io/2026/02/11/coinbase-agentic-wallets-ai-agents/) — a genuine independent-of-card-network product launch, though still Coinbase's own platform, not a third party building on top of it. See Coinbase Agentic Wallets — spend limits enforced server-side, native x402, and how thin the outside-usage evidence still is. - Adyen has announced "Adyen Agentic," an aggregator SDK supporting UCP, ACP, AP2, and Meta checkout in one integration (secondary source: https://www.digitalapplied.com/blog/agentic-commerce-standards-ucp-acp-ap2-2026-merchant-guide, citing Adyen's own positioning) — treat as company-claimed until a primary Adyen source or a live integration is found.

Quando si applica

when a claim names a "partner" without a repo link, a docs page, or a live checkout URL, that partner has not yet cleared this step. See Outside Implementations Tracker — separating shipped code from launch-partner press releases across the five agent-payment protocols.

Step 3: For Every Number, Find the Base (n, Window, Who Measured) — Refuse to Annualize a Single Observation

Any figure carrying a dollar amount, transaction count, rate, or volume must state the time window it covers, who measured it, and whether it is annualized (and from what shorter window). An annualized run rate is a projection, not a measured year.

- "Visa's stablecoin settlement hit $7 billion annualized run rate as of April 2026" (https://nai500.com/blog/2026/04/resilient-consumer-spending-boosts-visas-performance-with-annualized-stablecoin-settlement-volume-reaching-7-billion/): "annualized run rate" means Visa measured a shorter period and projected it to 12 months — a projection, not a measured 12-month total. The same source shows the run rate grew from $4.5B (January 2026) to $7B (April 2026) — quarter-over-quarter growth in *projected annual rates*, not in actual transactions measured over a year. Visa also expanded the pilot to nine blockchains as of April 2026 (Ethereum, Solana, Avalanche, Stellar, Base, Polygon, Canton Network, Arc, Tempo) in the same window. - x402 on-chain settlement — a worked example of a claim that fails this page's own method. A widely circulated figure gives 157,413,420 cumulative transactions as of 19 July 2026, attributed to agenteconomy.to (https://agenteconomy.to/stats/x402-transactions). Applying the procedure above, an adversarial refetch on 2026-08-24 could not reproduce it: the primary coverage checked (cryptobriefing.com, 17 August 2026) gives between 100 and 119 million total x402 transactions. The two do not reconcile, and no tracker publishes a method that would let a reader decide between them. Other counters disagree again — "200 million" per a June 2026 tracker whose own release could not be located, and a 30-day, not cumulative 75.41M snapshot on x402.org. The correct handling is the one this page teaches: do not pick the most convenient figure. State that the count is disputed, name each instrument and its window, and stop. A companion source (https://finance.yahoo.com/markets/crypto/articles/x402-foundation-activated-27-old-152440828.html) cites Artemis Analytics claiming over 95% of counted activity is infrastructure or test traffic rather than buyer-seller commerce — single-sourced, underlying data unpublished, so disputed rather than settled. Full reconciliation attempt in x402 'Tens of Millions of Transactions': Reconciling the Claim — a number without a base is not a fact. - Card interchange: Visa/Mastercard rates run roughly 1.40% + $0.05 to 2.4% + $0.10 per transaction as of 2026 (https://allaypay.com/blog/processing/current-interchange-rates-usa-updated-2026/, https://chargebacks911.com/mastercard-interchange-fees/); American Express's small-ticket category is 1.60% + $0.04 effective April 17, 2026 (https://merchantcostconsulting.com/lower-credit-card-processing-fees/2026-interchange-updates/). Every one of these has a rate class and an effective date — cite both, never just "card fees are ~2%." See Card Network Fees vs Micropayments — the interchange arithmetic, re-anchored to the primary schedules. - USDC on Base network transfer cost: $0.001–0.01 as of early 2026 (https://www.bitget.com/wiki/fees-for-transfering-usdc-on-the-base-network), achieved via Layer 2 transaction batching — this is a fee snapshot at a stated date, not a permanent constant.

Quando si applica

a figure with a dollar sign or a transaction count and no stated window, no stated measurer, and no note of whether it's annualized fails this step — flag it (unverified) and do not repeat it as fact.

Step 4: Distinguish "Launch Partner Named in a Press Release" from "Shipped Integration"

A launch partner is a commitment made at announcement time. A shipped integration is code in production accepting or initiating transactions using the protocol. The gap between the two is where vaporware lives. Evidence that actually distinguishes them: (a) the partner's own press release or developer docs naming the integration; (b) a public URL showing a transaction or checkout flow; (c) a GitHub repo with real implementation; (d) third-party audits or marketplace listings.

- ACP/ChatGPT Instant Checkout passes all four: OpenAI's own announcement, Stripe's own press release, published merchant-integration docs at https://docs.stripe.com/agentic-commerce/acp, and implied transaction evidence via live Etsy/Shopify storefronts. - AP2 launch partners: the September 2025 announcement lists partners, but there is no central status page tracking which have shipped; secondary sources describe Etsy's status inconsistently ("integrated" vs "in pilot") — the vocabulary itself signals unclear status. - Mastercard AP4M: announced June 10, 2026 with 31 named launch partners (Coinbase, Adyen, Stripe, Cloudflare, RippleX, Solana Foundation, Polygon, Aave Labs, OKX, MoonPay, and others per the press release). The phrase "announced protocol" signals it is not yet a shipped, production-tested infrastructure standard; developer docs exist but may be incomplete or reference-only, and evidence of merchant/wallet implementations outside Mastercard's own reference is scarce. - Visa Trusted Agent Protocol: described as "available today in the Visa Developer Center and GitHub" (https://developer.visa.com/capabilities/trusted-agent-protocol/overview, repo at https://github.com/visa/trusted-agent-protocol) — spec and reference code exist, but adoption by external merchants is not documented in the sources checked here.

Claim

"Visa's stablecoin settlement volume is annualized at $7 billion, proving widespread adoption of agent payments."

- Step 1 (spec): Visa Intelligent Commerce is documented at developer.visa.com, current status "in the process of development and deployment" — not a version-tagged published spec. - Step 2 (independent implementation): no public evidence of Visa Intelligent Commerce integrated by a third-party merchant or wallet outside Visa's own sandbox/demos; the $7B figure is a Visa-published number, not an independent measurement. - Step 3 (base): window is "annualized run rate" as of April 2026 — Visa measured a shorter period (weeks to a month) and projected to 12 months; measurer is Visa itself, not an independent auditor; scope is stablecoin settlement across nine blockchains. - Step 4 (partner vs shipped): Visa lists 100+ "partners" and 30+ "sandbox projects" with no central registry of which have production integrations; the $7B figure is Visa's own measurement on Visa's own rails.

Conclusion — partially verifiable with major caveats

Visa did announce the $7B annualized run rate in April 2026 (verifiable from the primary source). But it is a projection from a short window, not a measured 12-month total; it is not independently verified; it says nothing about adoption outside Visa's own ecosystem; and Visa Intelligent Commerce itself is still "in development," not shipped.

How to cite this correctly

"Visa announced an annualized run rate of $7 billion in stablecoin settlement volume in April 2026, measured on nine blockchains. This figure is Visa's own measurement projected from early-2026 run rates and does not yet reflect a full year of measured transactions or independent third-party verification."

Related

- Authorization Layer vs Settlement Layer — the split every agent-payment protocol page in this wiki must be read against — states the same four-part test as "the test for every other page in this wiki"; this page is the worked-out procedure behind that line. - Outside Implementations Tracker — separating shipped code from launch-partner press releases across the five agent-payment protocols — applies Step 2 and Step 4 systematically across all five protocols in this wiki, ranking them by strength of outside-implementation evidence. - x402 'Tens of Millions of Transactions': Reconciling the Claim — a number without a base is not a fact — the direct application of Step 3 to the "tens of millions of x402 transactions" claim, including the tension with the independent Bazaar enumeration on record in agent-economy. - Card Network Fees vs Micropayments — the interchange arithmetic, re-anchored to the primary schedules — Step 3's card-fee figures used and expanded with the full interchange arithmetic.