Mastercard Named Five Trust Pillars. Visa Scores 14. Mastercard.com Is Blocked.

Robauto Team · Published · Last updated

Mastercard’s Agent Pay Trust Framework lists identity, intent, controls, trusted execution, and intelligence. Live Robauto Agent Scans still show Visa at 14 and Stripe/Shopify at 0 of 5 commerce — while mastercard.com blocks the scan.

Mastercard named five trust pillars — Visa scores 14, mastercard.com is blocked

Mastercard just published five pillars for trusting AI agents with money — and a read-only Agent Scan of mastercard.com still hits a 403 bot wall. Live Robauto scores put visa.com at 14 with Agentic Commerce 0 of 5. The white paper asks the right questions. The open web has not answered them yet.

The Agent Pay Trust Framework’s five pillars

On September 30, 2026, Mastercard published Trust for agentic commerce, introducing the Agent Pay Trust Framework. Finextra’s October 1 interview with Ann Johnson (EVP, security solutions, Mastercard) spells out the same five layers:

PillarKey questionSource
Identity Who is acting? Mastercard
Intent What was authorized? Mastercard
Controls What is the agent allowed to do? Mastercard
Trusted execution Can the transaction be trusted? Mastercard
Intelligence Can risk be seen as agent behavior evolves? Mastercard

Johnson’s framing in the Finextra interview is blunt: trust determines scale. “Finding a pair of shoes through an agent is one thing, but trusting the agent to spend your money is another.”

Finextra also cites the scale forecasts Mastercard is working against:

That is the payments-network view of trust. Merchants still have to publish machine-readable surfaces those agents can discover on the open web.

Live Agent Scans, October 2, 2026

We ran fresh Robauto Agent Scans (~15:30 UTC). Agentic Commerce covers x402, ACP, UCP, AP2, and MPP:

DomainScoreAgentic CommerceNotesSource
mastercard.com blocked n/a 403 bot-management interstitial Agent Scan
visa.com 14 0 / 5 — Agent Scan
stripe.com 22 0 / 5 — Agent Scan
shopify.com 24 0 / 5 — Agent Scan
robauto.ai 98 5 / 5 verified Agent Scan
metareva.com 100 5 / 5 verified Agent Scan

Bar chart of Agent Scan scores: Mastercard blocked, Visa 14, Stripe 22, Shopify 24, Robauto 98, MetaReva 100

Networks and processors that shape how money moves still score in the teens and twenties on agent readiness, with commerce discovery at 0 of 5. Blocking a scanner is a legitimate ops choice — and it is also a decision about whether agents can see your trust story at all. Ceiling examples publish the full stack, including commerce discovery at 5 of 5.

Map the five pillars to what brands can ship

Issuers own identity tokens, intent proofs, and risk intelligence. Brands still win or lose on whether an agent can find honest, machine-readable surfaces without a human pasting a URL:

Mastercard pillarOpen-web surface brands can shipWhy it matters
Identity llms.txt, agent card / A2A profile, verified brand endpoints Agents need a stable “who you are” document, not only a homepage
Intent Clear product + policy pages; scoped MCP tools; AP2-aligned authorization hints Agents should know what a human authorized before they spend
Controls robots.txt allowlists, rate limits, MCP capability scopes, return/refund policies Hard limits beat soft promises when an agent acts at speed
Trusted execution UCP / ACP / x402 / AP2 discovery; checkout that agents can complete Commerce protocols turn “I found it” into “I can pay for it safely”
Intelligence Conversion pixel / analytics that distinguish agent traffic; audit logs You cannot manage agent risk you cannot see

A practical shipping order: Signal Basics first (robots, sitemap, llms.txt, readable HTML), then Agent Integration (MCP / agent card), then Agentic Commerce (x402, ACP, UCP, AP2, MPP). Trust papers raise the ceiling; discovery documents raise the floor.

{
  "as_of": "2026-10-02",
  "thesis": "trust_pillars_named_open_web_commerce_still_0_of_5",
  "mastercard_agent_pay_trust_framework": {
    "pillars": ["identity", "intent", "controls", "trusted_execution", "intelligence"],
    "source": "https://www.mastercard.com/us/en/news-and-trends/Insights/2026/mastercard-agentic-commerce-trust-white-paper.html"
  },
  "finextra_forecasts_cited": {
    "us_retail_agentic_commerce_2030_usd": "1_trillion",
    "b2b_agent_intermediated_2028_usd": "15_trillion_plus",
    "consumers_routinely_using_agents_to_buy_2030": "1_in_10",
    "source": "https://www.finextra.com/the-long-read/1774/how-banks-build-trust-in-agentic-commerce"
  },
  "brand_checklist": {
    "identity": ["llms.txt", "agent_card"],
    "intent": ["product_feed", "mcp_scopes"],
    "controls": ["robots.txt", "policies"],
    "trusted_execution": ["ucp", "acp", "x402", "ap2"],
    "intelligence": ["agent_aware_pixel"]
  },
  "example_scans": {
    "mastercard.com": "blocked_403",
    "visa.com": {"score": 14, "commerce": "0/5"},
    "stripe.com": {"score": 22, "commerce": "0/5"},
    "shopify.com": {"score": 24, "commerce": "0/5"},
    "robauto.ai": {"score": 98, "commerce": "5/5"},
    "metareva.com": {"score": 100, "commerce": "5/5"}
  }
}

Network pulse: agents crawl; referrals stay at zero

From live Robauto AI Search Data (updated 2026-10-02T15:00:11Z):

MetricValueSource
Total page views (90d) 1,335,762 AI Search Data
AI-classified views (90d) 307,518 AI Search Data
AI traffic share 23.02% AI Search Data
ai_referral visits 0 (0% of bot traffic) AI Search Data
llm_crawler share of bot traffic 53.5% (164,660 visits) AI Search Data
Agent Readiness Index median 20 / max 75 (79 scored, 20 blocked of 99 attempted) AI Search Data

Nearly a quarter of network traffic looks AI-classified, yet measured AI referrals are still zero. Crawlers read; agent-attributed visits that convert do not show up. Trust frameworks and open specs raise the ceiling for autonomous buying — the readiness median remains 20.

What to do next

Run a free Agent Scan on your domain. Map each Mastercard pillar to one shippable surface this week — start with llms.txt and an agent card, then layer commerce protocols. Guides live on Robauto Learn; watch crawls vs referrals on AI Search Data.

Sources

Where can an agent get this data live?

Robauto MCP — Agent Intelligence & Brand Discovery Infrastructure. A public, stateless Streamable HTTP endpoint at https://mcp.robauto.ai/mcp, callable with no key. Setup and the full tool list are in the Robauto MCP directory.

All blog posts