A Retailer’s Guide to AI Shopping Protocols: ACP, UCP, and MCP Explained

Key Takeaways

• AI traffic to Shopify grew 7× from Jan 2025 to early 2026
• MCP supplies structured product data; without it AI agents cannot query catalogs
• ACP enables ChatGPT checkout but redirects to merchant storefront for ownership
• UCP adds platform‑agnostic commerce layer with embedded checkout and dynamic payment negotiation
• Attribute completeness, not protocol adoption, is the biggest AI visibility bottleneck

Pulse Analysis

The rise of AI‑driven shopping assistants has turned product data into the new storefront. Protocols such as Model Context Protocol (MCP) act like the APIs of the mobile era, exposing inventory, pricing, and policy details in a machine‑readable format. When an AI like ChatGPT receives a query—"waterproof hiking boot under $180"—it queries MCP endpoints across merchants, filters results, and presents matches without a traditional product page. This shift means retailers can no longer rely on SEO or visual merchandising alone; they must treat their catalog as an API, ensuring every attribute from material to shipping window is clean and up‑to‑date.

Agentic Commerce Protocol (ACP) and Universal Commerce Protocol (UCP) sit atop the data layer, translating discovery into transaction. ACP, co‑developed by OpenAI and Stripe, enables checkout within ChatGPT but deliberately redirects shoppers to the merchant’s storefront, preserving brand relationships and data ownership. UCP, backed by Google and a coalition of major retailers, expands this capability across any AI surface, adding an Embedded Checkout Protocol and dynamic payment negotiation so merchants retain control over complex rules like discounts, taxes, and fulfillment options. Early adopters such as URBN, Etsy, and Coach have leveraged Stripe’s Agentic Commerce Suite to plug into ACP with minimal code changes, while Shopify’s default Agentic Storefronts already expose MCP to multiple assistants.

Despite the technical excitement, the real bottleneck is catalog richness. AI agents evaluate upwards of thirty structured attributes per product, yet many retailers expose only five to eight. Missing details—such as waterproof ratings, vegan certifications, or precise dimensions—cause the AI to filter the merchant out before a single recommendation is made. Tools like Paz.ai can audit attribute gaps across top‑selling SKUs, providing a prioritized roadmap for data enrichment. By fixing the data first, retailers unlock the full value of MCP, ACP, and UCP, turning AI assistants from a curiosity into a scalable sales channel.