Google introduces a new protocol for online shopping through AI agents

Since the emergence and expansion of artificial intelligence agents, many questions have arisen among users regarding the security involved in their implementation.

As we have explained in previous articles, an AI agent (Comet, Copilot, Operator) is capable of taking control of your device and performing actions independently, including making online purchases. To provide reassurance to users and to unify usage criteria, Google has developed the Universal Commerce Protocol (UCP), an open-source standard that establishes a common language and facilitates the purchasing process among consumers, eCommerce businesses, companies, and payment providers.

This initiative included collaboration from Shopify, Etsy, Wayfair, Target, and Walmart, and was supported by more than 20 global partners, including Adyen, American Express, Mastercard, Stripe, Visa, Zalando, among others.

What is the UCP

The UCP is a flexible, standardized protocol designed to benefit the entire commercial ecosystem:

• For businesses: The UCP allows them to showcase their unique products and services at key points of purchase via consumer interfaces such as AI Mode in Google Search and the Gemini app, among other future platforms.
• For AI platforms: They are able to simplify the onboarding of businesses via standardized APIs while maintaining flexibility to use MCP, A2A, and any preferred agent frameworks.
• For developers: The UCP is an open-source standard that can be enhanced and driven by the community.
• For payment providers: The open and modular design of the UCP payment manager enables open interoperability and allows for the selection of payment methods. Each authorization is backed by cryptographic proof of user consent.
• For consumers: It offers ease of purchase and security when making payments.

The UCP is designed to work with existing retail infrastructure and is compatible with the Agent Payments Protocol (AP2) to deliver secure support. Furthermore, it offers businesses flexible integration methods through API, Agent to Agent (A2A), and the Model Context Protocol (MCP).

Why a new standard is necessary

Users no longer shop in a linear fashion. They first ask questions, compare, research, and, when they are ready, want to complete their purchase without leaving the conversation. For this to work, brands must display real-time inventory, up-to-date pricing, and enable immediate payments within that same context.

The problem is that traditional technology is not prepared for this. Currently, every new platform or channel requires custom integrations, which creates immense technical complexity and hinders the evolution toward smarter and more automated shopping experiences.

The UCP removes this complexity by standardizing the entire commerce process (discovery, purchase, and order management) through a single secure layer:

• Unified integration: It reduces numerous integrations to a single point of entry.
• Shared language: It allows platforms and businesses to understand each other without needing specific adaptations.
• Extensible architecture: It adapts to new functions, experiences, and sectors.
• Security: It employs tokenized payments and verifiable credentials.

How the UCP works

Businesses publicize what AI agents are allowed to do (their capabilities, such as searching for products or making payments). The agents automatically discover these permissions and utilize them within the conversational experience.

For payments, UCP separates the consumer’s payment method from the payment processor, making it easier to work with multiple providers. It also allows for various types of connections (API, A2A, or MCP), depending on technical requirements.

As the infographic illustrates, on the left, the user initiates the interaction through tools such as Google Search (AI mode) or Gemini. On the right, businesses manage inventory, orders, and logistics, maintaining their role as the “registered merchant.”

In between, the Protocol serves as a “common language” enabling both parties to understand each other seamlessly. It is divided into two layers:

• Capabilities and extensions: Specific shopping process functions, such as the cart, identity linking, checkout, and order management.
• Underlying communication: The technical infrastructure that facilitates data exchange. This uses three primary routes: traditional APIs, the Model Context Protocol (MCP), and the Agent to Agent protocol (A2A).

Some of these capabilities are already developed, while others are in progress (those represented by dotted lines).

How to integrate UCP into your eCommerce

In Google’s official announcement, practical guidance is offered on how merchants can incorporate UCP into their platforms, by following these steps:

1. Configure a commercial server and add sample products to the store.
2. Prepare the enterprise server to accept requests from the agents.
3. Enable the agent to discover the commercial capabilities.
4. Invoke a payment function with the agent.
5. Apply discounts in the payment request with the agent.

While the example is quite technical, it demonstrates in HTML code format how a florist was able to incorporate an AI agent that discovered the business capabilities, initiate the payment process, and even apply a discount upon request.

“This tutorial demonstrates one of the capabilities that UCP offers. UCP also supports additional functionalities such as identity linking and order management, and it will continue to expand to deliver complete agent experiences for the consumer,” the company explains.

UCP integration with Google

It is important to clarify that the Protocol is provider-independent and can be utilized on any platform. UCP is not a Google product; it is an open standard. Google has simply been the first to put it into production and has applied it to its own conversational experiences.

Users can move from discovery to purchase within environments such as AI Mode in Google Search or Gemini, acquiring products directly from companies that meet the requirements.

When a person initiates a conversational query (for example, searching for a suitcase for a trip), Google displays shoppable products directly within that context. Payment is completed efficiently and securely using Google Pay, leveraging payment and shipping information already stored in Google Wallet.

This is illustrated in the following image:

To participate in Google’s implementation of UCP, companies must:

• Have an active Merchant Center account.
• Offer products that meet payment requirements.

In this way, Google can enable direct purchasing from inventory within its conversational experiences.

The UCP in Shopify Stores

Thanks to UCP, merchants using Shopify can sell directly within conversational experiences such as Google AI Mode and the Gemini app, managing everything from the Shopify Admin. In addition, certain merchants will participate in the Direct Offers pilot, which enables them to display exclusive offers within AI Mode precisely at the moment the user is ready to purchase, without leaving the conversation.

With the new Agentic plan, Shopify extends this model to all brands, including those that do not use Shopify as their online store. These brands may list their products in the Shopify Catalog, where data is standardized and enriched to be accurately displayed in AI-driven channels. By configuring information just once, products can appear in multiple environments such as ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini, thereby broadening brand reach and providing consumers with more purchasing options.

Image: Google.