The transition from a human-centric web to an agent-mediated digital economy represents the most significant shift in retail infrastructure since the inception of electronic commerce.
At the center of this transformation are two primary frameworks:
The Universal Commerce Protocol (UCP), championed by Google
The Agentic Commerce Protocol (ACP), developed by OpenAI
These protocols provide the standardized communication layers necessary for artificial intelligence systems (autonomous agents) to navigate the "Reasoning Web," a paradigm where software does not merely retrieve information but executes complex, multi-step transactions on behalf of users.
For professional peers in commerce architecture, and technical SEO, understanding the interplay between these protocols, the existing product data infrastructure (Google Merchant and Manufacturer Centers), and the evolving requirements of structured data is essential for maintaining visibility in the incoming era where AI agents act as the primary interface for consumption.
Architectural Foundations: Defining UCP and ACP
The Universal Commerce Protocol (UCP) and the Agentic Commerce Protocol (ACP) are designed to solve the problem of interoperability in a fragmented commerce ecosystem.
Before these standards, any AI platform wishing to facilitate a purchase would require bespoke integrations with every merchant’s backend, a model that does not scale across the millions of businesses globally.
UCP and ACP establish a "common language" that allows agents to discover products, evaluate options against user preferences, and complete checkouts through a unified set of APIs and primitives.
The Universal Commerce Protocol (UCP)
UCP is an open-source standard developed by Google, with significant contributions from partners including Shopify, Etsy, Wayfair, Target, and Walmart.
It is architected as a modular, extensible framework that covers the entire commerce lifecycle, from initial discovery and product comparison to checkout, order management, and post-purchase support.
The protocol is designed to be vendor-agnostic and surface-agnostic, meaning it can power interactions within Google Search AI Mode, the Gemini app, or any third-party agent that implements the standard.
The technical structure of UCP relies on a layered architecture:
The foundational layer is the Shopping Service, which defines core primitives such as checkout sessions, line items, and totals.
Above this reside Capabilities, which are standalone functional areas that a merchant can advertise, such as catalog access or returns processing.
The final layer consists of Extensions, which are domain-specific additions like loyalty programs or subscription management.
This modularity allows the protocol to evolve without breaking existing integrations, as an agent that does not support a specific extension simply ignores those fields during negotiation.
The Agentic Commerce Protocol (ACP)
ACP emerged from a partnership between OpenAI and Stripe to facilitate direct transactions within the ChatGPT environment.
While UCP aims for broad ecosystem governance, ACP is often characterized as a more specialized, high-velocity protocol optimized for conversational commerce.
ACP defines the sequence of structured, permission-based interactions between an agent and a merchant backend to perform commerce actions like buying, booking, or quoting.
It focuses heavily on the transaction moment, utilizing Stripe’s payment rails to handle delegated payment tokens, aka single-use, time-bound, and amount-restricted credentials that ensure the user remains in control.
Comparative Analysis: Commonalities and Divergences
UCP and ACP share several core philosophies that differentiate them from previous marketplace models.
Most critically, both protocols maintain the merchant’s role as the Merchant of Record.
Unlike aggregators that might sit between the customer and the business, UCP and ACP ensure that the merchant retains ownership of the customer relationship, fulfillment logic, and financial liability.
Furthermore, both protocols leverage structured product data as the "source code" of commerce, requiring real-time accuracy in pricing, inventory, and specifications.
Feature Dimension | Universal Commerce Protocol (UCP) | Agentic Commerce Protocol (ACP) |
|---|---|---|
Primary Backers | Google, Shopify, Walmart, Target | OpenAI, Stripe |
Philosophy | Decentralized, ecosystem-driven | Centralized, conversational-native |
Lifecycle Scope | Discovery to Post-Purchase | Transaction-focused execution |
Discovery Model | Distributed (published profiles) | Centralized (OpenAI listings/feeds) |
Payment Handling | Payment provider agnostic | Tightly integrated with Stripe/Tokens |
Transport Methods | REST, MCP, A2A, AP2 | REST, MCP |
Verification | Cryptographic proofs (AP2) | Shared Payment Tokens |
Despite these commonalities, the protocols diverge in their discovery and negotiation models.
UCP utilizes a decentralized "server-selects" architecture. A merchant publishes a capability profile at a standardized endpoint—the /.well-known/ucp JSON manifest—which declares everything an agent needs to know about supported services, payment handlers, and public keys. When an agent initiates contact, the merchant computes the intersection of supported capabilities and responds with a negotiated result.
ACP, conversely, has historically relied on a more centralized model where merchants apply directly to OpenAI and provide a product feed that is ingested into the ChatGPT ecosystem. While ACP is now moving toward greater interoperability, its origins remain deeply tied to the specific UX requirements of Large Language Models (LLMs) and conversational interfaces.
Data Infrastructure: Interactions with Google Merchant and Manufacturer Centers
For any commerce architecture, the efficacy of UCP or ACP implementation is entirely dependent on the quality of the underlying data.
In the Google ecosystem, this data is managed through the Merchant Center and Manufacturer Center, which together feed the Shopping Graph, which is a massive database containing over 50 billion listings that serves as the retrieval layer for AI agents.
Google Merchant Center (GMC) and the Agentic Feed
Google Merchant Center is the primary repository for offer-level data, including price, availability, and shipping information.
In the context of UCP, GMC is undergoing a transition from a static advertising tool to an active commerce layer.
Google, in fact, has introduced the native_commerce attribute, which is a mandatory field for products to be eligible for the "Buy" button in AI Mode and Gemini.
This attribute signals to the Google Shopping Graph that the merchant backend is equipped to handle UCP-compliant checkout requests.
Furthermore, Google is introducing dozens of new data attributes designed for conversational discovery.
These attributes extend beyond traditional keywords to capture nuances that agents require to make recommendations. This includes structured answers to common product questions, explicit lists of compatible accessories, and substitutes for products that may be out of stock. For a retailer, this means the feed must now include attributes that define "usage constraints" and "certifications," providing the context an AI needs to answer queries such as "Is this product safe for toddlers?" or "Is this charger compatible with a 2024 MacBook Air?"
Google Manufacturer Center and Product Truth
While the Merchant Center handles the transactional state, the Google Manufacturer Center is the authoritative source for product identity and specifications.
In an agentic environment, the Manufacturer Center acts as the "Product Truth" layer. When an agent evaluates multiple offers for the same product, it retrieves the authoritative specifications (dimensions, materials, technical specs) from the manufacturer's data to ensure consistency.
For brands, maintaining high-fidelity data in the Manufacturer Center is critical for AI Search, as it prevents agents from "hallucinating" product features or making incorrect compatibility claims based on low-quality merchant-provided snippets.
The Microsoft and Bing Equivalent Infrastructure
Microsoft’s ecosystem operates on a similar dual-track model, though it is more deeply integrated with the OpenAI ACP framework.
The Microsoft Merchant Center allows retailers to synchronize product feeds in the same XML/TSV formats used by Google, which are then used by Microsoft Copilot and ChatGPT Search.
The Bing ecosystem emphasizes the role of the Bing Webmaster Tools' AI Performance report, which provides insights into how products are being cited and surfaced within conversational queries.
While Microsoft does not have a separate "Manufacturer Center" with the same level of independent prominence as Google’s, it utilizes the centralized product feeds to bridge the gap between discovery in Bing Search and execution in Copilot Checkout.
Structured Data Strategy: Schema.org and the 2026 Paradigm Shift
Structured data, specifically in the form of JSON-LD, is the "machine-readable foundation" that allows AI agents to parse website content without visual scraping.
During the past months, Google introduced a significant consolidation of structured data requirements, retiring several specialized types (such as Book Actions and Automotive Listings) to prioritize core entity-based schemas that power the Knowledge Graph.
For commerce entities, the Product and Offer types remain the most critical components of search visibility and agentic eligibility.
The Role of WebMCP and Actionable Verbs
A major evolution in structured data is the emergence of the Web Model Context Protocol (WebMCP), which is described as the "new Schema.org moment".
If traditional Schema.org provides the "nouns" of the web (People, Places, Products), WebMCP provides the "verbs", aka the actionable tools that a website exposes to an agent.
Through WebMCP, a merchant can define actions like addToCart or bookAppointment as standardized interfaces, allowing the agent to "see" the functions available on a page and execute them directly via UCP or ACP pipes.
Critical Schema Mapping for Agentic Readiness
To participate effectively in UCP and ACP, merchants must ensure their on-page schema markup is not only valid but functionally rich.
The following table summarizes the essential mappings required for agentic commerce in the 2026 environment.
Schema.org Property | UCP/ACP Functional Use | Business Implication |
|---|---|---|
Product > gtin | Global Identity Resolution | Prevents offer duplication and ensures specifications are correctly mapped across platforms. |
Offer > availability | Real-time Transaction Eligibility | Agents will not recommend products that are not explicitly marked as InStock. |
Offer > price | Transaction Negotiation | Price mismatches between the agent’s quote and the final checkout result in failed sessions. |
MerchantReturnPolicy | Risk Evaluation Logic | Agents incorporate return ease into the user’s "best value" utility score. |
ShippingDetails | Fulfillment Calculation | Required for agents to provide a "final landed cost" within the chat interface. |
AggregateRating | Trust and Sentiment Scoring | Agents ingest reviews to summarize brand reputation and item quality for the user. |
The consistency between this on-page markup and the Merchant Center feed is paramount.
Agents are designed to cross-check these sources; if a PDP says an item is $49.99, but the feed says it is $59.99, the agent will flag the data as unreliable and likely exclude the product from the user's consideration set.
Deep Dive: The UCP Checkout State Machine
The technical core of UCP is the "Checkout State Machine," which governs the sequence of events from when a buyer expresses intent to when the transaction is finalized.
Unlike traditional checkouts, where a user is redirected to a website, the UCP session happens programmatically between the agent and the merchant server.
Session Creation and Update Endpoints
A UCP-compliant backend must implement three core REST endpoints:
Session Creation (POST /ucp/v1/checkout/create): The agent initiates a checkout session with the merchant, providing the SKU and quantity. The merchant server responds with a session ID and the current checkout state, including line items and estimated totals.
Session Update (POST /ucp/v1/checkout/update): As the agent adds shipping details or applies discount codes, the session is updated. The merchant server recalculates taxes, shipping costs, and promotions dynamically.
Session Completion (POST /ucp/v1/checkout/complete): Once the user confirms the purchase, the agent submits the payment token. The merchant processes the transaction and returns a final order confirmation.
The Two-Handshake Rule in Embedded Checkout
For merchants with complex product requirements (such as customization or bundling) that cannot be handled by a simple API, UCP supports an "Embedded Checkout" path using an iframe.
This involves a "Two-Handshake" security protocol.
First, the merchant's checkout UI performs a check-in with the host (e.g., Google Search) using postMessage.
If successful, the host may upgrade the communication to a private MessagePort, creating a secure, dedicated channel for the exchange of sensitive data like payment instruments.
This ensures that while the checkout feels native to the Google interface, the merchant remains the seller of record and handles all the underlying logic.
Case Study: Implementing Agentic Commerce for Asmodee’s Star Wars Legion
To move from theoretical protocol standards to actionable implementation, we examine the digital commerce footprint of Asmodee’s "Star Wars: Legion" miniatures game.
This category of wargaming is uniquely suited for agentic commerce because it involves high SKU complexity, compatibility constraints (expansions requiring base games), and a dedicated enthusiast base that often searches for specific, high-intent items.
Example 1: Collection Page Strategy
The collection page for Star Wars: Legion lists 91 products, including starter sets and operative expansions.
To ensure these products are discoverable and actionable by UCP-enabled agents, several steps must be taken to align the site’s technical infrastructure with the new protocols.
Actionable Tasks for Collection Visibility:
Implement Capability Discovery: Asmodee must host the /.well-known/ucp manifest declaring support for the dev.ucp.shopping.catalog capability. This allows agents to query the entire Legion collection programmatically rather than relying on visual scraping.
Structure Game Specifications: The visual labels for "2 Players," "90 Minutes," and "Ages 14+" should be mapped to additionalProperty within the Product schema. This enables an agent to answer a query like "Find me a Star Wars game for two people that takes about an hour and a half to play".
Sync Real-Time Availability: The collection page shows 17 items as "Sold Out". To prevent "Inventory Truth Breaks," Asmodee’s UCP endpoint must respond with a FAILED or UNAVAILABLE status if an agent attempts to initiate a checkout for these specific SKUs.
Example 2: Product Page Strategy
The "Imperial Shoretroopers" (SKU: SWQ199) is a specialized expansion set currently listed as sold out, with a reissue release date of February 20, 2026.
Actionable Tasks for Product-Level Conversion:
Attribute Mapping for Conversational Discovery: In the Google Merchant Center, Asmodee should use the new discovery attributes to define this product as an "Expansion" for the "Star Wars: Legion Core Set". This ensures that when a user asks an agent, "What do I need to play as Shoretroopers?", the agent can correctly identify both the expansion and the required base game.
Schema.org Refinement: The product page must utilize the Offer schema to explicitly state the release date (availabilityStarts) and the reissue relationship. As the product is a reissue of SWL41 with updated packaging, using the isVariantOf or successorOf logic in schema helps agents understand the product’s lineage.
Enable Agentic Checkout for Re-Stock: Once the February 2026 release date arrives, Asmodee should activate the native_commerce attribute in the Merchant Center feed. This will trigger the "Buy" button in Google’s AI surfaces, allowing users to pre-order or purchase the item using Google Pay stored in their Google Wallet.
The Road to Production: Merchant Readiness Checklist
For enterprise organizations, the adoption of UCP and ACP is an iterative process that requires alignment across engineering, product, and marketing departments.
Technical Readiness: The First 90 Days
Weeks 1–4: Foundations and Audits
Audit Product Schema: Use the Google Rich Results Test to ensure Product, Offer, and Organization schemas are valid and contain no errors.
Consolidate Data Sources: Ensure that product data fragmentations between marketing, operations, and IT are resolved to provide a single "source of truth" for the agentic layer.
Merchant Center Update: Begin populating the new conversational attributes (FAQs, compatibility, substitutes) in the Google Merchant Center.
Weeks 5–8: Protocol Implementation
Publish UCP Profile: Stand up the business server and host the manifest at /.well-known/ucp.
Configure Google Pay: Verify that your PSP supports Google Pay tokenization and set up the Merchant ID in the Google Pay & Wallet Console.
API Hardening: Secure the REST endpoints for session creation and updates to protect against malicious scrape-bots pretending to be commerce agents.
Weeks 9–12: Certification and Launch
Join the Waitlist: For UCP, submit the merchant interest form to Google for approval into the AI Mode and Gemini checkout experiences.
Apply for ACP: For non-Shopify merchants, apply directly to OpenAI for participation in ChatGPT Instant Checkout.
Establish Monitoring: Configure analytics to track "Agent Visibility" as a top-of-funnel metric, measuring how often your brand is cited or recommended by AI agents.
Economic Realities and Strategic Considerations
While the promise of agentic commerce is one of reduced friction and increased conversion, it is not economically neutral.
Retailers must navigate the "Platform Tolls" that accompany these new distribution channels. In the emerging model where platforms like ChatGPT or Gemini mediate the entire transaction, merchants may face transaction fees (e.g., the reported 4% in the ChatGPT-Shopify model) that sit on top of existing payment processing costs.
Furthermore, the loss of direct control over the storefront experience represents a significant strategic shift.
In fact, when a transaction occurs within Gemini, the merchant’s primary brand representation is the data they have provided to the Shopping Graph. This makes the "Business Agent" feature critical, as it allows brands to maintain a voice within the AI interface, answering questions and engaging consumers during high-intent moments.
The future of commerce is no longer a battle for keywords, but a battle for entities and relationships. By implementing UCP and ACP, retailers ensure that their products are not just visible to the human eye but actionable for the autonomous systems that will soon drive a quarter of all online retail sales.
The investment in high-fidelity structured data, protocol-ready backends, and real-time inventory synchronization is no longer a technical luxury but the foundational requirement for survival in the agentic era.
Article by
Gianluca Fiorelli
With almost 20 years of experience in web marketing, Gianluca Fiorelli is a Strategic and International SEO Consultant who helps businesses improve their visibility and performance on organic search. Gianluca collaborated with clients from various industries and regions, such as Glassdoor, Idealista, Rastreator.com, Outsystems, Chess.com, SIXT Ride, Vegetables by Bayer, Visit California, Gamepix, James Edition and many others.
A very active member of the SEO community, Gianluca daily shares his insights and best practices on SEO, content, Search marketing strategy and the evolution of Search on social media channels such as X, Bluesky and LinkedIn and through the blog on his website: IloveSEO.net.
stay in the loop