Testing whether AI shoppers can complete real buying tasks.
Agent-commerce QA
Agent-commerce QA measures whether AI assistants can discover the right product, understand merchant evidence, compare alternatives, and complete or hand off a purchase with confidence.
Merchant impact: This is the ShopperProof layer: run scenarios, find evidence gaps, apply fixes, and rerun to prove whether agent outcomes improved.
AI shopper win rateEvidence gapRerun
Source: ShopperProof methodologyThe share of prompts where an AI shopper chooses or completes the target store journey.
AI shopper win rate
AI shopper win rate is an outcome metric. It asks whether the agent selected the merchant, found enough evidence, and avoided competitor displacement for a realistic buying goal.
Merchant impact: A higher win rate means the store is easier for agents to recommend, cite, and route into checkout.
Agent-commerce QAPrompt stabilityCompetitor displacement
Source: ShopperProof report metricAn OpenAI and Stripe protocol for agent-assisted checkout and merchant order handoff.
Agentic Commerce Protocol (ACP)
ACP gives AI agents and merchants a shared way to coordinate purchase flows while keeping merchants in control of payments, systems, fulfillment, returns, and customer relationships.
Merchant impact: ACP readiness is not only an integration question. Product feed quality, variant facts, policy evidence, and checkout handoff clarity still determine whether agents recommend the merchant.
Instant CheckoutProduct feedPayment handoff
Source: OpenAI ACP announcementA Google and Shopify commerce protocol for agent discovery, negotiation, checkout, orders, and catalog capabilities.
Universal Commerce Protocol (UCP)
UCP lets merchants and agents publish profiles, negotiate supported capabilities, and continue shopping or checkout flows when a human handoff is needed.
Merchant impact: A Shopify merchant should think about discoverable capabilities, catalog facts, checkout continuation, fulfillment rules, and payment handler clarity.
Capability discoveryCheckout handoffAgent profile
Source: Shopify UCP architectureA Google protocol for proving that an agent is authorized to initiate a payment.
Agent Payments Protocol (AP2)
AP2 is an open payment protocol for agent-led transactions. It focuses on authorization, authenticity, accountability, and payment-method flexibility across users, merchants, and payment providers.
Merchant impact: AP2 readiness means the checkout path can distinguish user intent, agent authority, payment handoff, and merchant risk controls.
Payment handoffAuthorizationAgent-led payment
Source: Google Cloud AP2 announcementAn open protocol for agents to discover capabilities, exchange messages, and coordinate tasks.
Agent2Agent Protocol (A2A)
A2A defines how independent agent systems communicate through agent cards, messages, tasks, artifacts, streaming updates, and security schemes.
Merchant impact: For commerce, A2A matters when shopping, payment, support, fulfillment, or internal agents need to coordinate without sharing private tools or memory.
Agent cardTask lifecycleMCP
Source: A2A specificationA protocol for connecting models and agents to tools, data, and external context.
Model Context Protocol (MCP)
MCP standardizes communication between clients and servers that expose tools, resources, prompts, and contextual data to AI systems.
Merchant impact: A store or commerce platform may use MCP-like tooling to expose product data, order data, policies, or actions, but MCP alone does not prove a buyer scenario will succeed.
Tool accessProduct evidenceProtocol readiness
Source: MCP specificationA machine-readable profile describing an agent's identity, endpoint, skills, capabilities, and security requirements.
Agent card
In A2A, agent cards help other agents discover what an agent can do, what inputs and outputs it supports, and how to authenticate with it.
Merchant impact: Agent cards matter when merchants, apps, agencies, or commerce services expose specialized agents for shopping, support, fulfillment, or data retrieval.
A2ACapability discoveryAgent profile
Source: A2A agent discoveryStructured product data that agents and marketplaces use to discover items.
Product feed
A product feed usually includes titles, descriptions, images, prices, availability, variants, identifiers, categories, and other structured facts.
Merchant impact: Feeds need to answer the buying task, not just list SKUs. Missing fit, safety, compatibility, subscription, policy, or comparison facts can still make agents choose competitors.
Structured dataEvidence gapACP
Source: ShopperProof product evidenceMachine-readable page metadata, often schema, that describes products, offers, reviews, policies, or organization facts.
Structured data
Structured data helps search engines and AI systems parse store facts, but it must match visible storefront evidence and current product data.
Merchant impact: Structured data is a readiness input. ShopperProof checks whether agents can use it alongside page copy, policy pages, product data, and checkout signals.
Product feedEvidence gapSchema
Source: ShopperProof readiness guideA missing or unclear fact that prevents an agent from recommending or completing a purchase.
Evidence gap
Evidence gaps include missing size details, unclear ingredients, weak return terms, unsupported claims, absent compatibility data, or hidden delivery constraints.
Merchant impact: Fixing evidence gaps is usually merchant-controlled: update page copy, product metafields, policy pages, FAQs, reviews, or structured data, then rerun the same scenario.
Agent-commerce QARerunPrompt evidence
Source: ShopperProof audit reportsHow consistently repeated AI shopping prompts produce the same useful result.
Prompt stability
Prompt stability compares repeated runs of the same or similar buyer task to detect brittle recommendations, inconsistent winners, or low-confidence evidence.
Merchant impact: Low stability is a warning that the store may not be ready for customer-facing claims, benchmark reporting, or automated monitoring.
AI shopper win rateModel disagreementRerun
Source: ShopperProof stability labThe point where an agent moves a buyer from recommendation into checkout or a continuation flow.
Checkout handoff
A checkout handoff may be an embedded checkout, signed payment handoff, continue URL, cart transfer, or human-required step when the agent cannot complete everything by API.
Merchant impact: Handoff friction appears when the agent cannot preserve cart state, variant choice, payment context, shipping rules, or buyer intent.
UCPACPAP2
Source: Shopify UCP architectureWhen an AI shopper recommends another store because its evidence is easier to verify.
Competitor displacement
Competitor displacement can happen even when the merchant has a good product. The agent may prefer the competitor because size, policy, price, claims, reviews, or compatibility facts are clearer.
Merchant impact: ShopperProof reports show where the competitor won and which merchant-controlled facts could reverse the outcome on rerun.
AI shopper win rateEvidence gapFix backlog
Source: ShopperProof audit reports