Yesterday, Google’s CEO unveiled Universal Commerce Protocol (UCP) at NRF.  UCP represents a defining moment in the architecture of digital commerce; the strategic imperative is no longer merely about organizing the world’s information but about organizing the world’s commercial intent and empowering merchants to leverage their own data to construct superior customer experiences.  This shift is not incremental; it is a fundamental re-platforming of the digital economy, where Google is uniquely positioned to serve as the orchestrator within a “virtuous cycle” of interaction among retailers, consumers, and intelligent agents. 

“For many people, discovery is the fun part of shopping. Making a decision is where things get harder. As an indecisive shopper myself, I’m looking forward to the day when agents can help me get from discovery to purchase.

At Google, we’re busy laying the groundwork for this agentic ecosystem to work well. That includes building a common language for these systems and services to talk to each other.

As a next step we are introducing the Universal Commerce Protocol (UCP), designed for the era of agentic commerce. It was built to meet the needs of retailers AND customers, keeping the full customer relationship front and center — from the moments of discovery to decision and beyond”. – Sundar Pichai NRF – Jan 11 2026

While entities like OpenAI are engaged in a race to build better cognitive models, Google created UCP to enable a new economy. By establishing the standard for how agents discover, negotiate, and transact, Google becomes the commercialization infrastructure of data collaboration, further expanding the collaborative ecosystem it has cultivated over two decades.

Google’s Gravity Enables UCP

Google is the world’s largest data business. While most of us understand how consumer services like Gmail and Maps (4 billion consumers) and the Android operating system (4.2 billion surfaces), provide deterministic data on intent. Few outside of the advertising world see the massive signal ingestion engines where 3rd party data is provided to Google to measure, optimize and tune ad effectiveness (see Blog Data Games).

The scale of Google’s virtuous cycle is hard to comprehend, a “data gravity” that pulls merchants into the ecosystem. Unlike fragmented open-web approaches, Google offers a unified view of the consumer across the entire journey: from the initial spark of interest in Search or YouTube to the final in a SKU level sales receipt provided by merchant. Historically the data exchange has been a mash up of proprietary services and data transfers. UCP will standardize and transform “top of the funnel” data, turning passive signals into active commercial mandates (ex, intent) that enable fragmented systems to discover and negotiate capabilities, all with alignment to consumer and merchant objectives. 

UCP Accelerates the Flywheel

UCP accelerates the platform by addressing fragmentation in the agentic. For example, a single user appears as different entities across their phone, desktop, laptop, and TV.  UCP, anchored by Google’s unified identity has the ability to resolve this into a single, continuous identity.

This resolution powers a flywheel effect:

1. Merchant Data Injection: Merchants contribute SKU-level data and inventory real-time feeds via UCP, motivated by the promise of high-intent traffic.
2. Agent Intelligence: This data enriches the “Specialist Agents” running on Gemini and Google Cloud, making them more capable and helpful to consumers.
3. Consumer Engagement: Consumers, finding friction-free and highly relevant agentic experiences, consolidate their commercial activity on Google surfaces.
4. Network Optimization: The increased transaction volume feeds back into Campaign Manager 360 (CM360), optimizing ad spend and proving ROI, which in turn drives further merchant investment.

From “Data Games” to Discovery, Engagement and Action

As discussed in Data Games the tension between platforms and merchants centered on the fear of commoditization and data leakage.  Merchants were wary that sharing data with a platform would empower a competitor or dilute their direct customer relationship. Google built the trust of merchants in tools like Campaign Manager 360, which optimizes campaigns across both Google and non-google properties.  As one top retailer CMO said “We share data with Google because it drives incremental sales across all advertising and enables us to optimize all of our ad spend”.

Today, the eCommerce Funnel is full of custom data flows across organizations and platforms. The CMO manages an ad budget that delivers consumers to the home page, the eCom head designs the CX to convert this intent to a purchase, the payment team measures risk and fraud.  

Historically, the data flow between merchants and Google occurred primarily within the Merchant Center, often facilitated by intermediaries like LiveRamp. In this model, merchants would upload customer lists (hashed emails, phone numbers) to create “Customer Match” segments for targeting on Search, YouTube, and Gmail.

• Mechanism: This relied on “Passthrough Activation” or “Enhanced Matching,” where PII (Personally Identifiable Information) was hashed and matched against Google’s user graph.
• Limitations: While effective for retargeting, this model was constrained by privacy regulations (GDPR, CCPA), the latency of batch processing, and the inherent risk of moving sensitive customer data between environments. It was a mechanism of the “Web 2.0” era—static, list-based, and inherently leaky.

UCP Standardizes Roles and Flows

Agentic commerce collapses the funnel, the retailer’s data is needed to provide a better search but there is no “search marketing” or commercial construct, the retailers eCommerce UI is not needed.  UCP brings standardization to the data flows, and roles necessary for the evolution of commerce. 

• Collapsing the Funnel: UCP removes the friction of “landing pages” and navigation. By allowing agents to transact directly against a merchant API product feed, the traditional awareness-to-purchase funnel collapses into a single conversational turn.
• Data as the New SEO: Retailer data is no longer just for analytics; it is the fuel for discovery. Without providing accurate, real-time SKU data (inventory, price, fulfillment), a retailer is invisible to the Agent. This is an enhancement to MCP
• Evolution of “Search Marketing”: There is no commercial construct for “buying keywords” in an agentic dialogue. UCP replaces the Search Engine Results Page auction with a relevance-based matching engine, forcing a complete rethink of “search marketing” budgets.
• Headless Commerce: The retailer’s eCommerce UI is rendered unnecessary for these transactions. UCP standardizes the presentation layer, meaning the merchant provides the logic and fulfillment, but Google provides the interface (Native) or a standardized container (Embedded).
• Standardized Data Flows: UCP brings rigid standardization to previously chaotic data flows. It aligns product definitions, checkout sessions, and return loops into a common protocol that every agent and merchant can speak.
• Evolution of Roles: UCP clarifies the necessary evolution of commerce roles. The Merchant becomes the Inventory & Risk Owner; Google becomes the Platform & Signal Provider; and the Agent becomes the Consumer Interface. This specialization allows each party to optimize their specific value contribution to the economy.

UCP Brokers Needs and Capabilities (Tech Dive)

UCP brokers interactions between consumer needs (expressed through AI agents) and business capabilities (inventory, pricing, and checkout) by establishing a standardized, machine-readable “common language” that decouples the agent from the specific backend infrastructure of the merchant. An “MVC” framework for commerce.

1. Dynamic Capability Discovery via JSON Manifests To match an agent’s need with a merchant’s ability to fulfill it, UCP requires businesses to publish a standardized JSON manifest located at /.well-known/ucp.

• • The Mechanism: This manifest acts as a directory of the business’s supported services (e.g., Shopping) and specific “capabilities” (e.g., checkout, product discovery, discounts).
  • The Result: Agents do not require hard-coded integrations for every merchant. Instead, they query this manifest to dynamically discover endpoints, supported features, and payment configurations, allowing them to instantly understand how to interact with a specific business.

2. Standardization of Functional Primitives UCP breaks complex commerce flows down into “functional primitives”—core building blocks that define what a business can do.

• • Core Capabilities: Primitives such as “product discovery” and “checkout” are standardized so that any agent can invoke them regardless of the underlying platform (e.g., Shopify, Salesforce).
  • Extensions: To handle specific business logic, UCP uses an “extensions framework.” For example, a “discount” extension can augment the standard checkout capability, allowing the agent to apply coupon codes dynamically during the transaction request.

3. Flexible Transport Bindings (A2A, MCP, API) UCP brokers the actual communication of data by supporting multiple transport protocols, giving businesses flexibility in how they expose their capabilities to agents.

• • Interoperability: A specific capability, such as Checkout, can have a REST API binding or a Model Context Protocol (MCP) binding depending on the platform’s needs.
  • Seamless Integration: This allows agents to communicate using the Agent2Agent (A2A) protocol or standard APIs, ensuring that the “need” (the user’s request) can be transmitted securely to the “capability” (the business backend) using the most appropriate technical pipe.

4. Modular Payment Architecture To bridge the gap between a user’s wallet and a merchant’s processor, UCP models a unique payment architecture that separates the “instrument” from the “handler”.

• • Instruments vs. Handlers: It distinguishes between what consumers use to pay (instruments like credit cards or digital wallets) and the payment processors (handlers like Stripe or Adyen).
  • Provable Payments: This modular design allows the agent to present the correct payment method while ensuring the transaction is routed to the correct backend processor, backed by cryptographic proof of user consent (via compatibility with AP2).

5. Contextual Continuity By standardizing these interactions, UCP allows the agent to maintain “conversational context” throughout the funnel. The protocol enables real-time inventory checks and dynamic pricing to occur within the search or chat interface, collapsing the traditional N x N integration bottleneck into a single abstraction layer that supports the full lifecycle from discovery to decision.

Google Cloud as the Modeling Engine

Google Cloud (GC) serves as the environment where this collaboration and modeling occur. It is the foundry for the new economy. By co-locating the data (via BigQuery Omni and Analytics Hub) and the compute (Vertex AI), Google minimizes latency and maximizes security.

• Specialist Agents: These are not generic LLMs. They are highly tuned models that understand the specific nuances of a retailer’s catalog. For example, a Home Depot agent built on this stack understands “rough-in valves” not just linguistically, but commercially—knowing which SKUs are in stock locally and offer the best margin.
• Objective Control: The merchant defines the objective function. Unlike an Amazon algorithm which optimizes for Amazon’s ecosystem yield, a UCP agent on GCP optimizes for the merchant’s defined outcome (e.g., maximize profit, clear excess inventory, acquire new customers).

This shift from “Merchant Center lists” to “GCP collaborative modeling” is the technical foundation that allows Google to claim it is “enabling merchants to use their own data” rather than coopting it. Internally, Google’s AP2 can serve as the accounting engine that tracks what agent delivered which value for any given interaction. 

AP2 – Risk Signals and the Liability Shift

As I outlined last week in AP2 as Signals, Google understands that existing merchant owned transaction risk will dominate the early days of agentic. While AP2 and associated mandates are capable of supporting a liability shift, there are too many moving parts to make this work in near term. To be clear AP2 is technically elegant and Visa/MA have in place both the technology and rules (DAF/TAF) to enable a liability shift, the politics with Issuers, Merchants, Networks and Big Tech are not aligned.

UCP provides for 2 payment flows Native and Embedded. To support the diverse technical maturity and strategic needs of the merchant ecosystem, UCP offers two distinct integration paths: Native and Embedded. Both leverage Google’s identity and payment infrastructure but offer different levels of control and friction.

The Native integration represents the “pure” agentic vision. It is powered by a Headless GPay implementation that combines credentials across Google Pay and Chrome Autofill (CAF),

• Mechanism: The merchant implements a REST API. The agent builds the cart and, upon user confirmation, Google sends a payment token directly to the merchant’s API,
• Network Tokenization: Within the card rails, this is powered by Network Tokenization Services. The merchant receives a dynamic token (DPAN) rather than a raw PAN, ensuring security and improving authorization rates.
• User Experience: The user experience is “headless” where there is no redirect, no checkout form to fill. The transaction happens conversationally or in the background.
• Strategic Fit: This modality is ideal for high-trust, high-frequency interactions where friction reduction is paramount.

For merchants who require more control or support for diverse payment methods, the Embedded Checkout option is available.

• Mechanism: The merchant provides a URL to a web-based checkout UI, which Google embeds in an iframe within the agent interface.
• Merchant Control: “You will implement a checkout UI that will be embedded in the checkout flow through an iframe”. This ensures the merchant retains control over the pixels, the terms of service presentation, and the upsell logic.
• Third-Party Wallets: Crucially, this modality allows for the leverage of third-party wallets like PayPal and Paze.  A merchant can offer a Paze button within the iframe, and the user can checkout using their bank-backed wallet.
• Credential Facilitation (PAR/COF): In the embedded model, Google can facilitate the use of credentials stored within the merchant’s own system. By passing a PAR (Payment Account Reference), hashed email, hashed PAN, or other stable identifier, Google enables the merchant to recognize the user and trigger their existing Card on File (COF) flow.11 This bridges the gap to the merchant’s existing loyal customer base, ensuring high authorization rates without requiring a new payment instrument.

Within the embedded flow, Google explicitly lists the risk signals provided it intends to provide within UCP. 

1. Session Data: Context on the user’s interaction session (e.g., duration, interaction type).
2. Device Data: Telemetry from the Android/Chrome surface. This may include a chrome “software secure element” ID.
3. Browser Data: Signals from the Chrome instance.
4. Shipping & Billing Data: Verification of address consistency.
5. Payment Verification: Confirmation of the token assurance level.

These signals are designed to replace the lost device fingerprinting data. By consuming these signals, merchants can confidently approve agentic transactions even without a formal liability shift from the bank.

UCP Enables a New Economy

The acceleration of the Google platform via UCP is not merely a technical upgrade; it is an economic restructuring. By unifying the surfaces (Android, Search, Maps) with a common commercial protocol, Google has created a network effect that is virtually impossible to replicate.

• For the Consumer: It delivers the “magic” of conversational commerce—agents that know what you want, find it instantly, and buy it securely without forms or friction.
• For the Merchant: It offers the only viable path to “high-intent” agentic traffic. While they must bear the risk in the near term, the “Data Games” have shown that the ROI of Google’s network justifies the cost. The shift to Private Join and Compute resolves the data sovereignty concerns that previously held them back.
• For the Advertiser: CM360 provides the only closed-loop attribution system that spans the open web and the walled garden, powered by the granular SKU-level data that UCP unlocks.

Google is not just participating in the agentic economy; it is building the operating system for it. While OpenAI works to build a better model, Google has built the economy; a network with commercial constructs at both ends, where retailers and consumers engage in a virtuous cycle of value creation.