Inside Google’s UCP and OpenAI’s ACP race to own AI checkout

Google’s Universal Commerce Protocol (UCP) and OpenAI’s Agentic Commerce Protocol (ACP), which powers ChatGPT Instant Checkout have emerged as notable changes in how digital commerce may work across web platforms, search engines, and AI interfaces.

For years, online shopping has depended on users navigating traditional storefronts, adding items to carts, and checking out on merchant websites. Increasingly, users now ask AI assistants for product recommendations, comparisons, and purchase options — behavior that has exposed a gap between intent and transaction in digital experiences.

UCP and ACP are both responses to this gap. They create standards that let external systems handle commerce flows, from product discovery to payment confirmation in a consistent, machine-understandable way.

Neither feature is yet a finished, mass-market product, but both signal a shift in how merchants and retailers approach online buying experiences. Both protocols help reduce the distance between discovery and purchase by letting users complete transactions within the interface where product discovery happens. This is a departure from legacy e-commerce patterns that rely on navigating to separate shop sites.

However, they differ in scope, ecosystem integration, and technical design, and may point to diverging visions for how AI will influence commerce going forward. Advertisers and developers alike will want to understand how these mechanisms function, how they compare to existing commerce systems, and what the early strategic implications are for digital retail.

What is Google’s Universal Commerce Protocol?

Google’s Universal Commerce Protocol (UCP) is an open standard designed to let AI agents discover merchant capabilities, invoke checkout flows, and manage orders across the entire shopping journey. Instead of checkout flows being tightly owned by each individual site, UCP aims to introduce a common way to represent product details and order actions so that different interfaces can execute commerce requests in the same structured way.

UCP is not a standalone product; it is the plumbing that enables secure and structured commerce between users, agents, and merchants. It standardizes how merchants expose product data, inventory, capabilities (like discounts or returns), and checkout endpoints so that multiple agent systems can handle the same information consistently.

The idea is that if merchants expose their product catalogs and checkout endpoints through UCP, agents outside of the merchant site, like Google Search or AI assistants, could initiate a purchase without needing a bespoke integration for every partner.

This is different from earlier schemes, where every integration required building a custom API for each merchant or platform. According to Google, UCP was co-developed with partners including Shopify, Etsy, Wayfair, and Target.

Google has already rolled out “in-chat” purchase options powered by UCP in AI Mode and Gemini in the U.S., letting some users complete purchases directly inside those experiences without navigating to external retailer sites. Early participating merchants include Etsy and Wayfair, with further integrations planned for Shopify, Target, and Walmart.

UCP is intended to address some of the limitations of traditional search-based shopping by enabling a deeper understanding and reasoning about commerce relationships. For example, because the protocol feeds transaction and fulfillment data back into Google’s Shopping Graph, Gemini can leverage years of structured commerce relationships to answer complex shopper queries, such as personalized device recommendations, and complete those flows in context.

Under UCP, retailers remain the Merchant of Record; they set pricing, manage fulfillment, and retain customer data even if the sale completes inside an AI interaction.

Salesforce recently announced support for UCP in its Agentforce Commerce platform, helping merchants provide native checkout and discovery experiences on conversational AI surfaces, including secure payment methods through Google Pay.

How UCP works in practice

When a user queries Google in AI Mode or interacts with Gemini about a product, the system:

• identifies merchants whose UCP-enabled listings match the query
• checks available payment handlers published by merchants
• surfaces a purchase option inline with search or conversational results
• completes the transaction through a supported payment method like Google Pay without a redirect or external checkout
  

This integrates commerce into the search experience itself, turning discovery into fulfillment in a single flow.

OpenAI’s Instant Checkout: Agentic Commerce Protocol 

OpenAI introduced Instant Checkout in ChatGPT in late 2025, making it possible for users to discover products and buy them entirely within a conversation. Early support includes purchases from Shopify and Etsy sellers.

Instant Checkout uses the Agentic Commerce Protocol (ACP), an open standard co-developed by OpenAI and Stripe. The protocol enables AI agents and businesses to work together to fulfill orders, transmit payment information securely, and let buyers complete transactions without leaving the app.

The flow for Instant Checkout using ACP works like this: after a user asks ChatGPT for product recommendations, the system ranks products across the web based on inventory, pricing, and availability. If a product supports Instant Checkout, it shows a ‘Buy’ button directly in chat. With a tap, users can confirm shipping and payment details stored in their profile, and the purchase is completed without leaving the conversation.

With ACP:

• Merchants can make product catalogs discoverable inside ChatGPT
• Users can proceed from browsing to purchase within the chat without being redirected
• Payments are executed through supported processors such as Stripe, Apple Pay and Google Pay
• Discovery and checkout flows are standardized in a way that allows agents to handle the transaction end-to-end

ACP initially focuses primarily on checkout-centric interactions, letting users complete purchases inside chat flows. The protocol also makes merchant products discoverable organically when they match user queries, converting discovery into direct purchases.

Unlike UCP, which is baked into Google’s own conversational surfaces, Instant Checkout is built into ChatGPT’s chat interface and sits on Stripe’s ACP standard, making it available to merchants who expose product feeds and enable checkout through Stripe’s APIs and shared payment tokens. The protocol is endorsed by payments partners, including PayPal.

Reports indicate that merchants pay fees to OpenAI for Instant Checkout, but the exact structure and rate are not officially disclosed by OpenAI. Independent merchant reporting cites potential fees of around 4% of the transaction.

Integrations are also emerging beyond Etsy and Shopify, with partnerships enabling deeper commerce experiences. For example, Instacart enables grocery shopping with full checkout inside ChatGPT by linking inventory and delivery flows into the conversation.

How these two systems relate

Google UCP and OpenAI’s ACP address similar goals: reducing friction between purchase intent and transaction completion. Users can go from finding a product to buying it without leaving the interface where the search or recommendation happened.

UCP and ACP are open or collaborative protocols. UCP encourages broad ecosystem participation, while ACP is designed to let any compatible agent and merchant interact once integrated through shared standards. Both systems aim to provide protocols that multiple players can adopt, lowering technical barriers for merchants.

In both ecosystems, merchants remain the merchant of record, handling fulfillment, returns, and customer relationships. The AI interface simply orchestrates the flow.

Where UCP and ACP diverge

While Google’s Universal Commerce Protocol (UCP) and OpenAI’s Agentic Commerce Protocol (ACP) are often grouped together as “AI shopping infrastructure,” they reflect two different strategic visions for how commerce should function in AI environments.

Google’s UCP is designed as a broad, foundational layer for agentic commerce. It is not just about checkout. Instead, it aims to standardize the entire lifecycle of an AI-mediated transaction — from product discovery and availability checks to secure checkout sessions, payments, and even post-purchase actions like confirmations and updates. In other words, UCP is attempting to become a universal commerce language that any AI agent, retailer, or payments provider can speak. It is infrastructure-oriented and ecosystem-wide by design.

ACP, by contrast, is more tightly focused on enabling purchases inside conversational AI interfaces, specifically within ChatGPT. While it is open and extensible, its current implementation centers on solving a more immediate problem: how to allow users to move from recommendation to completed purchase inside a chat experience without redirecting to an external site. In that sense, ACP is transaction-centric. It is optimized for in-conversation checkout rather than architecting a full universal commerce stack across multiple AI systems.

The ecosystems backing each protocol also reflect this difference in ambition and positioning. UCP is being advanced by Google in partnership with large retailers, payment networks, and commerce platforms. It is integrated directly into AI Mode in Search and the Gemini app, signaling Google’s intent to weave commerce into its broader AI and search infrastructure. The goal appears to be cross-platform interoperability — a standard that works wherever AI agents interact with merchants.

ACP, meanwhile, is closely tied to OpenAI’s product roadmap and Stripe’s payment infrastructure. Its early rollout through ChatGPT’s Instant Checkout demonstrates a product-led approach: start with a high-usage AI interface, integrate payments deeply, and expand merchant adoption outward. Rather than positioning itself as a universal commerce backbone across the web, ACP currently operates as a commerce layer within ChatGPT’s ecosystem.

Technically, UCP is architected to support multiple transport mechanisms and protocol extensions, allowing different agent systems and merchant platforms to integrate using shared specifications. ACP is more tightly integrated with Stripe’s payment primitives and tokenization framework, reflecting its origin as a collaboration between an AI platform and a payments company.

The distinction, then, is strategic. UCP is attempting to define how AI-driven commerce works everywhere. ACP is focused on making AI-driven checkout work inside one of the world’s most widely used conversational AI products, with the potential to expand from there.

Still, their coexistence illustrates a broader evolution: commerce experiences are becoming loosely coupled from specific storefronts and can live in search results, assistant interfaces, or contextual agents.