Powering Product Discovery in ChatGPT
ChatGPT integrates the Agentic Commerce Protocol (ACP) to enable visual, interactive product discovery, side-by-side comparisons, and direct merchant integration—enhancing e-commerce capabilities at the agent level.
Shopping Experience Upgraded via Agentic Commerce Protocol
ChatGPT has officially deployed the Agentic Commerce Protocol (ACP), transforming traditional conversational search into a structured, visually immersive product discovery workflow. ACP is not a standalone model but an agent-level protocol layer standardizing LLM-to-ecommerce-service interactions. It supports dynamic catalog retrieval, real-time price/inventory API calls, multi-dimensional attribute parsing (e.g., size, material, user reviews), and cross-platform price comparison.
Key Capabilities in Production
- Product Discovery: Users describe needs in natural language (e.g., “lightweight waterproof hiking shoes for outdoor use, under ¥1200”), and ChatGPT generates candidate lists with images, parameter cards, and purchase links;
- Side-by-Side Comparison: Up to 4 products can be compared simultaneously in a dedicated sidebar panel across key dimensions (brand, price, rating, delivery time, return policy), with median response latency <800ms;
- Merchant Deep Integration: Official APIs from Shopify, Walmart Marketplace, and Amazon Seller Central are integrated, enabling order status lookup, automatic coupon application, and post-purchase support navigation.
Technical Positioning and Boundaries
ACP does not involve new model training; it operates atop GPT-4-turbo (gpt-4-turbo-2024-04-09), leveraging its function calling and multimodal (vision + text) understanding capabilities. The protocol specification is open-sourced on GitHub (repository: openai/agent-commerce-protocol), and v0.3 mandates that all merchant adapters pass the OpenAI ACP Compliance Test Suite v1.2.