Agentic Commerce Protocol: AI Shopping Agents Guide
The Agentic Commerce Protocol (ACP) standardizes how AI shopping agents discover, compare, and purchase products. Setup guide for merchants.
Agentic Commerce Market (2025)
Projected Market Size (2033)
CAGR Growth Rate
Shoppers Using AI Agents by 2030
Key Takeaways
Online shopping is about to undergo the most significant structural change since the introduction of mobile commerce. AI agents are learning to shop on behalf of consumers, not just recommending products but autonomously discovering, comparing, negotiating, and purchasing them. The Agentic Commerce Protocol (ACP) is the emerging standard that makes this possible by defining how AI shopping agents communicate with merchants at the protocol level.
The numbers behind this shift are substantial. The agentic commerce market reached $547 million in 2025 and is projected to grow to $5.2 billion by 2033 at a 32.5% compound annual growth rate. Morgan Stanley estimates that roughly half of all online shoppers will use AI agents for purchasing decisions by 2030. This is not a speculative future scenario. Shopify, Google, Stripe, Walmart, and Amazon are already building infrastructure for it. For merchants already exploring how Shopify's agentic commerce and Google UCP integration work in practice, ACP provides the broader protocol context that connects all these pieces together.
What Is the Agentic Commerce Protocol
The Agentic Commerce Protocol is an open standard that defines the rules of engagement between AI shopping agents and merchants. Think of it as HTTP for commerce: just as HTTP standardized how web browsers request and receive web pages, ACP standardizes how AI agents request product information, compare options, and execute purchases. Without a shared protocol, every agent would need custom integrations with every merchant, creating an unsustainable matrix of point-to-point connections.
ACP covers the entire transaction lifecycle. It specifies how agents discover merchant catalogs, how they query for products matching user preferences, how real-time pricing and inventory availability are communicated, how carts are created and managed, and how checkout and payment authorization are orchestrated. Each of these interactions follows a defined schema that both agents and merchants implement, enabling any ACP-compliant agent to transact with any ACP-compliant merchant.
Agents query structured catalog endpoints to find products matching user intent, preferences, budget constraints, and brand affinities without scraping web pages or parsing HTML.
Programmatic cart creation, item addition and removal, coupon application, and shipping option selection through machine-readable API calls rather than UI interaction.
Secure payment authorization, address verification, order confirmation, and receipt generation handled through protocol-level interactions with built-in fraud prevention.
The protocol is designed to be extensible. Core interactions are mandatory for compliance, but merchants can expose additional capabilities like loyalty program integration, subscription management, bundle pricing, or personalized offers through optional protocol extensions. This allows the ecosystem to evolve without breaking backward compatibility, a lesson learned from earlier attempts at standardizing eCommerce APIs that failed because they were either too rigid or too permissive.
How ACP Works: Discovery to Checkout
Understanding the ACP transaction flow reveals why this protocol matters for merchants. A typical agentic commerce interaction involves six distinct phases, each with defined request and response schemas that both parties implement.
Merchant Discovery
Agent queries the ACP registry or merchant-published discovery endpoints to find stores selling the target product category. Returns merchant capabilities, supported payment methods, and catalog metadata.
Catalog Query
Agent sends structured product queries with filters for price range, specifications, availability, brand, ratings, and other attributes. Merchant responds with matching products in a normalized schema.
Real-Time Pricing
Agent requests current pricing including dynamic discounts, bundle offers, loyalty rewards, and shipping costs for the user's location. Prices are guaranteed for a defined validity window.
Cart Assembly
Agent creates a cart with selected items, applies available coupons or promotions, selects shipping method, and calculates the final order total with taxes and fees.
Payment Authorization
Agent initiates payment through the user's pre-authorized payment method (often via Stripe MPP or similar protocol). Transaction is verified against user-defined spending limits and merchant restrictions.
Order Confirmation
Merchant confirms the order, provides tracking information, and returns a structured receipt. Agent stores the order details and can monitor fulfillment status on the user's behalf.
The entire flow can complete in seconds. A user tells their AI agent “order me a replacement HEPA filter for my Dyson V15” and the agent handles merchant discovery, product matching across multiple stores, price comparison, and purchase execution autonomously. The user receives a confirmation with the order details, total cost, and expected delivery date.
What makes this fundamentally different from existing comparison shopping tools is the closed-loop execution. Traditional price comparison sites show results but hand the user off to the merchant's website for checkout. ACP-powered agents complete the entire transaction without the user ever visiting a merchant site. This has profound implications for how merchants think about customer acquisition and conversion, a point we will explore in the eCommerce solutions context.
AI Shopping Agents: ChatGPT, Perplexity, Gemini
The agent side of the agentic commerce equation is evolving rapidly. Three major platforms have launched or are building shopping agent capabilities that will consume the ACP, UCP, and MPP protocols. Understanding their approaches reveals where the market is heading and which agent behaviors merchants need to optimize for.
OpenAI integrated product search and comparison directly into ChatGPT conversations. Users describe what they need in natural language, and ChatGPT returns product recommendations with images, prices, ratings, and purchase links. The system pulls from multiple merchant data sources and is expanding toward full checkout capability.
Perplexity launched direct purchasing within its search interface. Pro subscribers can buy products found through search queries without leaving Perplexity. The system handles checkout, payment, and order tracking natively, making it one of the first fully closed-loop AI shopping experiences available to consumers.
Google is connecting Gemini to its existing Google Shopping and Merchant Center infrastructure. Gemini's shopping capabilities leverage Google's massive product database, merchant relationships, and UCP feeds. The integration with Google Pay creates a seamless path from product discovery to payment.
The competitive dynamics between these platforms will shape how ACP evolves. ChatGPT has the largest conversational user base. Perplexity has the most aggressive commerce integration with actual purchasing built in. Gemini has the deepest merchant data through Google Shopping and the most mature payment infrastructure through Google Pay. Each is incentivized to support open protocols like ACP because it expands their product catalog coverage. For merchants already thinking about how to optimize for these AI assistants, our guide on ChatGPT as a shopping assistant for eCommerce covers practical optimization strategies.
Beyond these three major platforms, specialized shopping agents are emerging for specific verticals. Fashion agents that understand fit and style preferences, grocery agents that optimize for dietary restrictions and meal planning, electronics agents that compare technical specifications across hundreds of products. These vertical agents often provide better results than general-purpose ones because they encode domain-specific knowledge about what attributes matter most for each product category.
Merchant Setup and AI-Discoverable Product Feeds
For merchants, the most urgent question is practical: what do you need to do to make your products visible to AI shopping agents? The answer involves three layers of technical implementation, each building on the previous one.
Implement JSON-LD Product schema markup on all product pages. Include detailed attributes: price, availability, brand, SKU, GTIN, condition, shipping details, and review aggregates. This is the foundation that agents use for initial product matching.
Expose a product catalog API that supports filtering, sorting, and faceted search. Include real-time inventory levels and pricing. This is what agents call for catalog queries beyond basic discovery, enabling precise product matching against user requirements.
Implement API endpoints for cart creation, item management, coupon application, shipping calculation, and order placement. This is the critical layer that enables agents to complete transactions without redirecting to a browser-based checkout flow.
Register with ACP discovery services and agent platforms. Submit your merchant profile, catalog capabilities, supported payment methods, and geographic coverage. This is how agents find your store when users ask for products you sell.
Merchants on major platforms have a significant advantage. Shopify is building ACP and UCP compatibility into its core platform, meaning merchants on Shopify will gain agent discoverability through platform updates rather than custom development. WooCommerce, BigCommerce, and Magento are developing similar integrations. Independent merchants running custom eCommerce implementations will need to build these capabilities themselves or adopt middleware that bridges their existing systems to the ACP standard.
Priority action for merchants: Start with structured product data. Even before full ACP adoption, JSON-LD Product markup makes your products visible to AI agents that crawl and index the web. ChatGPT Shopping, Perplexity, and Gemini all consume schema.org Product markup today. This single step provides immediate visibility while the full protocol stack matures.
For businesses evaluating the full scope of AI-driven eCommerce automation beyond product feeds, our guide on AI eCommerce automation tools for 2026 covers the broader toolkit including inventory management, pricing optimization, and customer service automation.
Security, Trust, and Transaction Verification
Giving an AI agent the ability to spend money on your behalf raises legitimate security questions. The agentic commerce ecosystem addresses these through multiple overlapping security layers, each designed to prevent a different category of abuse or error.
Mastercard developed the Verifiable Intent framework specifically for AI agent transactions. It cryptographically links the user's original request to the agent's purchase action, creating an auditable chain of intent that proves the agent acted on genuine user instructions.
Users define spending limits, approved merchant categories, maximum per-transaction amounts, and geographic restrictions. The agent cannot exceed these constraints regardless of what it infers from user intent. Rules are enforced at the payment protocol level, not by the agent itself.
ACP requires merchants to verify their identity through the protocol registration process. Verified merchants receive cryptographic certificates that agents validate before initiating transactions, preventing fake stores from intercepting agent purchases.
Mastercard's Verifiable Intent deserves particular attention because it solves a problem unique to agentic commerce: proving that a purchase was intentional. When a human clicks “Buy Now,” the click itself serves as proof of intent. When an AI agent initiates a purchase, there is no equivalent human action at the moment of transaction. Verifiable Intent creates a cryptographic record that traces from the user's original instruction (for example, “order me more coffee pods when I run low”) through the agent's reasoning process to the specific purchase it executes. This chain is verifiable by payment processors, banks, and merchants.
The practical implication for consumers is that agentic commerce transactions will have stronger fraud protection than traditional online purchases. Every agent transaction has a documented intent chain, spending constraint enforcement, and merchant identity verification. A human typing a credit card number into a phishing site has none of these protections.
For merchants: Transaction verification requirements will increase for ACP-participating stores. Ensure your payment processing supports the additional metadata fields that Verifiable Intent and MPP require. Work with your payment processor to enable agent transaction acceptance early, as there may be verification lead times.
Impact on Traditional eCommerce Funnels
The most disruptive aspect of agentic commerce is what it does to the traditional eCommerce conversion funnel. The funnel that merchants have spent two decades optimizing, from awareness to consideration to conversion, fundamentally changes when AI agents mediate the purchase decision.
Awareness: SEO, ads, social media drive traffic to site
Consideration: Product pages, reviews, comparison convince buyer
Conversion: Cart, checkout, payment completed by human
Retention: Email marketing, retargeting bring repeat visits
Discovery: Agent queries ACP endpoints, not your website
Evaluation: Agent compares data across merchants in seconds
Transaction: Agent completes purchase via API, no human UI
Repeat: Agent remembers preferences, auto-reorders
The implications cascade through every aspect of eCommerce strategy. Website design and UX become less important for agent-mediated transactions because the agent never sees your website. Brand storytelling and emotional marketing have reduced impact because agents evaluate on data attributes, not feelings. Product photography and lifestyle content matter less when the agent is comparing JSON specifications rather than browsing visual product pages.
What matters more in an agentic commerce world is data quality and completeness. The merchant with the most comprehensive, accurate, and up-to-date product data wins the agent's recommendation. Competitive pricing becomes even more transparent because agents compare across all available merchants simultaneously. Fulfillment speed and reliability become stronger differentiators because agents can factor delivery track records into their recommendations. Review scores and return rates carry more weight when an agent aggregates them mathematically rather than a human scanning them casually.
This does not mean traditional eCommerce optimization becomes irrelevant. Roughly half of consumers will continue shopping manually, and high-consideration purchases like furniture, electronics, and luxury goods will remain browse-heavy for years. But the merchant who ignores the agentic channel is leaving an increasingly large share of transactions on the table. The businesses exploring AI and digital transformation strategies today will be positioned to capture this shift as it accelerates.
Preparing Your Business for Agentic Commerce
The transition to agentic commerce is not a distant event to plan for later. The protocols are being finalized now, the major platforms are integrating them now, and early-adopting merchants are gaining positioning advantages now. Here is a prioritized action plan for businesses at any stage of readiness.
- •Add comprehensive JSON-LD Product schema markup to all product pages
- •Ensure product data is complete: pricing, availability, SKU, GTIN, brand, specifications
- •Submit and maintain Google Merchant Center feeds with real-time inventory
- •Audit and improve product review collection for higher aggregate scores
- •Build or enable a machine-readable product catalog API with filtering and search
- •Implement real-time pricing endpoints that return current prices with validity windows
- •Enable headless checkout capabilities that support API-driven order placement
- •Contact your payment processor about MPP and agent transaction support timelines
- •Register with ACP discovery services and agent platforms
- •Implement full ACP endpoint compliance: discovery, catalog, cart, checkout
- •Enable Verifiable Intent and MPP payment acceptance
- •Set up analytics to track agent-originated transactions separately from human traffic
For businesses on Shopify, the timeline accelerates significantly because Shopify is building protocol compliance into the platform. Shopify merchants should monitor Shopify Editions announcements for ACP and UCP features, enable them as they become available, and focus their effort on data quality rather than technical implementation.
Independent merchants and those on platforms without native ACP support should evaluate middleware solutions or consider migrating to platforms with built-in agent commerce capabilities. The cost of custom ACP implementation is significant, and the ROI depends on your product category. High- volume, commodity, and repeat-purchase categories will see agentic commerce adoption fastest. Niche, luxury, and experience-driven categories will see it later.
Conclusion
The Agentic Commerce Protocol represents the infrastructure layer of a fundamental shift in how products are discovered, evaluated, and purchased online. Combined with Google's Universal Commerce Protocol for product data exchange and Stripe's Machine Payments Protocol for autonomous payment authorization, ACP creates a complete stack for AI-mediated commerce. The market numbers, $547 million today growing to $5.2 billion by 2033, reflect real investment and adoption, not speculative projections.
For merchants, the action items are clear and sequential: structured product data first, machine-readable APIs second, full protocol compliance third. The merchants who treat agentic commerce as a future consideration rather than a present priority will find themselves invisible to an increasingly significant share of purchasing activity. The protocols are being finalized, the platforms are integrating them, and the consumers are adopting the agents. The window to establish positioning in the agentic commerce ecosystem is open now but will not stay open indefinitely.
Ready for Agentic Commerce?
Making your products discoverable to AI shopping agents requires structured data, machine-readable APIs, and protocol compliance. Our team helps eCommerce businesses build the technical foundation for agentic commerce.
Related Articles
Continue exploring with these related guides