Evidence discipline

Agentic commerce: myths versus evidence

This market moves quickly enough that marketing language can turn a scoped feature into a universal promise. The distinction matters to merchants, developers and buyers.

Myth: one protocol makes every checkout agent-ready

Evidence: UCP, ACP, AP2, MCP, x402 and trust protocols address different layers. A merchant must still validate its own catalog, checkout, consent, payment and support paths.

Myth: a public API proves production agentic-commerce support

Evidence: an API documents a possible integration surface. It does not prove that a merchant has enabled it, passed conformance tests or supports every buyer context. This site records such distinctions in the Observatory.

Myth: product discovery and payment authorization are the same problem

Evidence: discovery determines what an agent can evaluate. Payment authorization determines whether a buyer has approved a specific spend. Treat them as separate controls even when they appear in one user journey.

Myth: a bot identifier is proof of an authorized agent

Evidence: a user-agent string is an assertion. Visa’s Trusted Agent Protocol describes cryptographic message signatures, signed identity context and replay protection for the specific trust model it documents.

Myth: an announcement means universal availability

Evidence: eligibility, geography, access tiers, feature flags and merchant configuration remain relevant. State the observed scope and source date, then re-test before presenting a feature as live.

Myth: a high readiness score is a certification

Evidence: the readiness diagnostic and public signal scanner are transparent decision aids. They cannot attest to security, production transaction volume or approval from a protocol owner.