Universal Commerce Protocol (UCP): the open standard redefining commerce in the AI agents era

Executive summary

On January 11, 2026, at NRF Retail’s Big Show in New York, Google CEO Sundar Pichai unveiled the Universal Commerce Protocol (UCP) — an open standard designed to enable AI agents to manage the entire purchase journey, from product discovery to after-sales service. Co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, and backed by more than 20 major players including Visa, Mastercard, Stripe, Adyen, American Express, and Zalando, UCP is not merely a technical protocol: it is Google’s bid to become the universal checkout infrastructure for the entire AI ecosystem.

With McKinsey projecting an agentic commerce market of $3 to $5 trillion by 2030, and 60% of searches now “zero-click” according to Bain & Company, UCP arrives at a pivotal moment when the traditional purchase journey is being fundamentally questioned.

What is agentic commerce?

Agentic commerce refers to a model in which AI agents autonomously complete purchases on behalf of users. This is no longer simply an LLM recommending a product — the agent is now the gatekeeper of the transaction itself: it browses, compares, negotiates, and buys on behalf of the consumer.

This paradigm fundamentally transforms the relationship between consumer and merchant. As McKinsey puts it: “This is not simply an evolution of e-commerce — it is a complete reimagining of what it means to shop.” Instead of a linear journey (search → browse → compare → buy), shopping becomes a continuous intent-driven flow, where the agent navigates the ecosystem on the human’s behalf.

The N × N problem

The current AI commerce landscape suffers from a problem of exponential complexity. Every AI agent (Gemini, ChatGPT, Copilot) must build a specific integration with every retailer (Shopify, Walmart, Etsy). And conversely, every retailer must maintain separate connections with every AI platform. This is the “N × N” problem — a multiplication of integrations that makes current conversational commerce fundamentally broken.

UCP solves this by creating a common language: a single integration to communicate with any compatible agent. This is what industry observers call “the HTTPS moment of retail” — just as HTTPS standardized secure web communication, UCP standardizes secure AI commerce.

Technical architecture of UCP

Design philosophy: layered protocols

UCP draws lessons from TCP/IP by separating responsibilities into modular layers rather than building a monolithic protocol. As Shopify’s engineering team, co-architect of the protocol, explains: “Monolithic protocols eventually collapse under complexity — too rigid to adapt, too slow to evolve. Carefully layered protocols survive and thrive.”

The architecture breaks down into three levels:

• Shopping Service: defines the core transactional primitives — checkout session, line items, totals, messages, status
• Capabilities: add major functional domains (Checkout, Orders, Catalog), each versioned independently
• Extensions: enrich capabilities with domain-specific schemas via composition

The six core capabilities

The concrete implementation of UCP rests on six capabilities covering the entire commercial lifecycle:

Atomic primitives and interoperability

UCP relies on four atomic primitives: `ucp_discovery`, `ucp_negotiate`, `ucp_identity`, and `ucp_checkout`. Each can be executed independently and asynchronously — an agent can negotiate with twenty merchants simultaneously without starting twenty separate sessions.

The protocol is designed to be compatible with existing standards:

• Agent2Agent (A2A): inter-agent coordination
• Agent Payments Protocol (AP2): secure payment via mandates and verifiable credentials
• Model Context Protocol (MCP): context and service discovery
• OAuth 2.0: standard authentication and authorization

Security model: privacy-first

UCP implements a data sovereignty model where sensitive user information (address, payment methods) remains in their personal “agentic vault.” The merchant only receives a Right-to-Pay token verified via the `ucp_identity` primitive, which drastically minimizes the risk of data leaks. Payment is handled via tokenization, allowing merchants to use their existing payment integrations.

UCP vs ACP: two protocols, two ecosystems

Google’s UCP does not arrive in a vacuum. In September 2025, OpenAI and Stripe launched the Agentic Commerce Protocol (ACP), the first open standard for agent-driven commerce. The two protocols coexist but serve different surfaces.

The fundamental distinction: while ACP focuses on the checkout → fulfillment → payment funnel, UCP covers the entire commercial cycle with its six capabilities. Google explicitly aims to become the universal checkout infrastructure for all the AI ecosystem, rather than another walled garden.

The strategic recommendation for most serious merchants: implement both protocols, as they serve different AI surfaces. Walmart already supports ChatGPT (ACP ecosystem) and Gemini (UCP ecosystem), and Shopify merchants automatically access both via platform-level integrations.

The ACO concept: Agentic Commerce Optimization

A new optimization paradigm

The emergence of UCP and ACP creates a new optimization field that SEO experts are beginning to formalize under the acronym ACO — Agentic Commerce Optimization. As a Search Engine Journal SEO expert puts it: we are moving from “optimizing for the click” to “optimizing for the selection.”

This is a fundamental shift: AI agents do not browse pages the way humans do. They query APIs, parse product feeds, and evaluate structured data. Optimization must therefore target a dual audience — human and agent — simultaneously.

The four pillars of ACO

1. Crawlability: agents still follow links, navigate paths, and understand information architecture 2. Format: content must be concise, without filler, but with enough unique value and consistency across the entire site 3. Structured Data: agents are increasingly dependent on existing standards, particularly open-source standards 4. Brand Authority and Sentiment: without positive brand sentiment, convincing an agent to cite a brand during discovery remains a major challenge. Third-party perspectives (reviews, mentions) carry increasing weight in agent grounding

Schema.org: the glue of the ecosystem

As Pascal Fleury of Google stated at Google Search Central Live in December 2025: “Schema.org is the glue that binds all these ontologies together.” UCP handles the transaction; schema.org helps agents decide with whom to transact.

The product schema must be comprehensive: `name`, `description`, `SKU`, `GTIN`, `brand`, images, `offers` (including `price`, `priceCurrency`, `availability`, `seller`), `aggregateRating`, `review`, `shippingDetails` with delivery estimates. The `Organization`, `FAQPage` schemas, and GS1 data for granular product detail complete the technical arsenal.

Technical implementation guide

Step 1: Prepare the Merchant Center Feed

UCP uses the existing Merchant Center feed as a discovery layer. Beyond on-site schema, several elements are critical:

• Return policies (mandatory for Merchant of Record status): costs, return windows, links to policies. For advanced accounts, each sub-account requires its own policies
• Customer support information: structured contact data, potentially usable by agents to resolve level-1 queries
• Agentic checkout eligibility: add the `native_commerce` attribute to the feed — without this attribute, products are not eligible for checkout in AI Mode
• Product identifiers: every product must have an ID correlated to the ID used in the checkout API
• Consumer warnings: any product warning must use the `consumer_notice` attribute

Google recommends using a supplemental data source in Merchant Center rather than modifying the primary feed, to avoid formatting errors.

Step 2: Optimize conversational commerce attributes

Google has announced dozens of new Merchant Center attributes designed for discovery in the era of conversational commerce, on surfaces like AI Mode, Gemini, and Business Agent. These attributes go beyond traditional keywords:

• Frequently asked questions answers: product and brand FAQs integrated into the feed
• Compatibility: compatible accessories, up-sell opportunities
• Substitution: replacement products to handle out-of-stock situations
• Related products: cross-sell opportunities
• Granular details: beyond “purple,” specify “dark purple” or even the specific shade name (e.g., “Wolf”). Same logic for sizes, materials, and compositions

Step 3: Implement UCP

The technical implementation follows a structured process:

1. Join the UCP waitlist via Google Merchant Center 2. Configure shipping, returns, and product feed for discovery and purchasing on Google surfaces 3. Choose supported capabilities: UCP allows selecting the subset of capabilities and extensions relevant to your business 4. Choose the communication medium: REST APIs, MCP, or A2A depending on development needs 5. Google validation: the integration must be approved before going live on AI Mode and Gemini

For Shopify merchants, integration is largely automated via “agentic storefronts” — UCP is natively enabled, you simply need to verify the configuration. Wayfair and Etsy are already live with UCP for direct purchasing in AI Mode since February 11, 2026.

Step 4: Embedded Checkout Protocol (ECP)

When a handoff to checkout is required, the agent loads the embedded checkout via a `continue_url`. The protocol establishes a bidirectional JSON-RPC 2.0 channel:

• The merchant sends state updates
• The agent provides credentials and context
• Payment collection uses the host’s native payment sheet
• Address selection draws from the agent’s wallet

This system derives from Shopify’s Checkout Kit, distilled into an open protocol built on years of large-scale embedded checkout operation, with PCI v4 compliance and advanced sandboxing.

Multi-modal fan-out and agent selection

How agents select products

A key concept in UCP optimization is multi-modal fan-out — the process by which an agent decomposes a query or image into multiple search intents to evaluate which attributes to prioritize.

For example, from a single product image, a visual fan-out simulator like WordLift’s can decompose the image across three “horizon” depths, revealing the selection paths the agent might take: from the specific brand (Horizon 1) to the broad category (Horizon 2: “performance running gear”) and then to adjacent products (Horizon 3: sports equipment in general).

This analysis shows the breadth of LLM consideration during selection, and how conversational attributes can be structured to maximize the chances of being selected during fan-out.

Social proof and third-party perspective

Regardless of the quality of on-site optimization, data integrity must be validated by trusted third-party sources. Platforms like Trustpilot and G2 are frequently cited and trusted by most LLMs — positive brand and product reviews remain essential for satisfying the consensus agents seek before initiating any transaction.

The "zero-click" impact: the existential crisis of e-commerce

The visit-purchase decoupling

UCP accelerates what experts call the “zero-click” crisis of e-commerce. When users can complete purchases directly in AI interfaces, they never click through to the merchant’s website — which fundamentally challenges the e-commerce traffic model that has existed for 25 years.

The numbers speak for themselves: according to Bain & Company (February 2026), 60% of searches are now “zero-click,” and 58% of consumers use AI tools to research products — but only 17% feel comfortable finalizing a purchase through an AI interface. This gap between research and execution represents both a massive risk and opportunity.

Who controls the shopper?

With UCP, the merchant remains the Merchant of Record — retaining control of their customer data and commercial relationships. But the discovery surface and purchase journey are operated by AI. The e-commerce site is no longer the primary entry point to commerce; product data becomes the shared language between merchants and agents.

This shift raises existential strategic questions: marketplaces, loyalty platforms, and intermediaries that thrive on human-driven engagement will need to decide whether to build their own agents, whether to welcome or block agent traffic, and how to monetize when algorithms — not people — make purchasing decisions.

Market projections and sizing

The economic opportunity of agentic commerce is the subject of converging projections from major consulting firms:

McKinsey compares the potential impact of agentic commerce to the web and mobile revolutions, arguing that the transition could unfold even faster this time, as AI systems “borrow the rails” of existing commercial infrastructure rather than waiting for new ones to be built.

UCP roadmap: beyond retail

The protocol plans significant expansion in the coming months:

• Multi-item carts and complex baskets: beyond single-item checkout, with bundling, promotions, tax/shipping logic
• Loyalty and account linking: standardized management of loyalty programs, cross-merchant points redemption and member pricing
• Post-purchase support: order tracking, returns, customer service handoff
• Personalization signals: enriched cross-sell/upsell, wishlists, history, contextual recommendations
• New verticals: expansion beyond retail into travel, services, digital goods, and food & beverage

UCP already supports recurring payments via redundant authorization patterns — the user authorizes their agent to manage recurring charges under specific conditions (e.g., pay up to $50/month for electricity, replenish essentials when stock drops below a threshold).

New Google tools for agentic commerce

Business Agent

Alongside UCP, Google is launching Business Agent — a branded AI agent that shoppers can query directly in Search. This agent functions as a virtual salesperson capable of answering product questions in the brand’s voice.

Eligible retailers can activate and customize this agent in Merchant Center, with expanding capabilities: training on merchant data, customer insights, related product offers, and direct purchasing including agentic checkout. Lowe’s, Michael’s, Poshmark, and Reebok are among the early partners.

Direct Offers

Direct Offers is a new Google Ads pilot allowing advertisers to present exclusive offers (e.g., -20%) directly in AI Mode to shoppers ready to buy. Google uses AI to determine when an offer is relevant to display. The pilot initially focuses on discounts, with planned expansion to bundles and free shipping. Petco, e.l.f. Cosmetics, Samsonite, and Rugs USA are among the pilot brands.

Action plan: prepare now

Priority 1: Immediate actions

1. Join the UCP waitlist via Google channels 2. Audit Merchant Center: complete return policies, add the `native_commerce` attribute, verify product identifiers 3. Audit schema.org: complete Product schema (SKU, GTIN, offers, aggregateRating, shippingDetails, FAQPage) 4. Test agent perception: ask AI platforms to describe your products — verify visibility, accuracy, and completeness

Priority 2: Conversational optimization

5. Populate conversational attributes: Q&A, compatibility, substitutions, related products 6. Refine descriptive granularity: move from “purple” to “dark purple” / exact shade for all attributes (color, size, materials) 7. Develop social proof: intensify review collection on Trustpilot, G2, and trusted third-party platforms

Priority 3: Multi-protocol strategy

8. Also implement ACP if relevant (Shopify merchants access it automatically) 9. Train the technical team on UCP documentation and code samples available on GitHub 10. Set up monitoring: track agent selections, conversions via AI surfaces, and sentiment in AI responses

Who should prioritize UCP?

• Complex catalogs: UCP’s full-journey approach handles discovery better for large inventories
• Post-purchase as competitive advantage: UCP standardizes returns, exchanges, and support
• Google Search traffic: UCP powers AI Mode, where the majority of product queries still begin
• Existing Google Shopping participants: existing Merchant Center data accelerates UCP integration

The verdict

UCP is not a fleeting LLM gimmick — it is a structural component of the biggest shift in e-commerce since the advent of mobile. As one expert puts it: “This is moving faster than most previous commercial shifts — brands that wait for full deployment signals will already be behind.”

The central message is clear: ACO = SEO. The fundamentals — structured data, quality content, brand authority, social proof — have not changed. What has changed is that the audience consuming and acting on these signals is no longer exclusively human. The AI agent is now a first-class citizen in your optimization strategy, and UCP is the language it speaks to transact.

References

• Google partners Walmart, Shopify and others for AI-powered shopping — Google CEO Sundar Pichai unveils Universal Commerce Protocol at NRF 2026.
• New tech and tools for retailers to succeed in an agentic shopping era — Overview of Google’s new open standard for agentic commerce.
• Google announces a new protocol to facilitate commerce using AI agents — TechCrunch coverage of the UCP announcement.
• McKinsey forecasts up to $5 trillion in agentic commerce sales by 2030 — McKinsey projects $1 trillion in US online retail by 2030.
• Agentic Commerce: Why AI is Winning the Search War but Losing the Checkout Battle — Bain & Company (Feb 2026): 60% of searches are now zero-click.
• UCP vs ACP: Which Agentic Commerce Protocol Should Retailers Choose? — Comprehensive comparison of UCP and ACP with implementation guidance.
• Agentic Commerce and the End of Search in Retail — Analysis of agentic commerce reshaping the retail landscape.
• Google UCP Explained: The 2026 Guide to AI-Powered Shopping — Detailed guide to UCP and its 20+ major partners.
• Universal Commerce Protocol (UCP): Google’s AI Shopping Revolution Explained — How UCP solves the fragmentation problem in AI commerce.
• Building the Universal Commerce Protocol (2026) — Shopify Engineering — Shopify’s technical deep-dive on co-developing UCP.
• UCP vs ACP: Which Standard Will Rule the Agentic Web in 2026? — Comparative analysis of the two leading protocols.
• Google’s Universal Commerce Protocol — What It Changes — Lengow’s analysis of UCP’s implications.
• Universal Commerce Protocol — ucp.dev — Official UCP specification.
• Google Universal Commerce Protocol (UCP) Guide — Google’s official developer documentation.
• UCP vs ACP: Which Protocol Does Your Store Need? — Practical comparison for merchants.
• Overview — Google Universal Commerce Protocol (UCP) Guide — Google’s official implementation overview.
• Preparing Your Shopify Store for UCP — Guide to activating UCP on Shopify agentic storefronts.
• Google UCP & OpenAI ACP — Agentic commerce checklist — How to prepare for UCP and ACP in 2026.
• What Retailers Need To Know About Agentic Commerce — McKinsey’s $5 trillion projection and merchant data infrastructure.
• UCP Code Samples — Official Google code samples for UCP implementation.