Agentic commerce is buying in which an AI agent discovers, compares and purchases products for a person. Agents rely on structured product feeds and open protocols such as OpenAI's Agentic Commerce Protocol, Google's Universal Commerce Protocol and Agent Payments Protocol. Preparation starts in the PIM: identifiers, complete attributes, variants, current price and availability.
Agentic commerce is shopping in which an AI agent searches, compares and, with permission, buys products on a person's behalf. Agents cannot read a product page the way a person does, so they depend on structured product feeds and open protocols. Product data teams prepare by making identifiers, attributes, variants, price and availability complete, consistent and current.
What is agentic commerce?
An AI shopping agent is software built on a large language model that takes a goal ("find white running shoes under 120 euros in size 44") and carries out the steps: discovery, comparison, cart and checkout. Between September 2025 and May 2026, the major AI and commerce platforms published protocols that define how agents, merchants and payment providers exchange data.
The product data requirement is constant across them: an agent can only recommend and buy what it can identify and compare.
Which protocols and programmes matter?
| Protocol or programme | Announced | Backers | Role |
|---|---|---|---|
| Agentic Commerce Protocol (ACP) | 29 September 2025 | OpenAI, Stripe | Product feeds, checkout and delegated payment for ChatGPT |
| Agent Payments Protocol (AP2) | 16 September 2025 | Google, more than 60 organisations | Signed Intent and Cart Mandates proving what the user authorised |
| Universal Commerce Protocol (UCP) | 11 January 2026 | Google with Shopify, Etsy, Wayfair, Target, Walmart | Discovery to order management, compatible with AP2 |
| Shopify Catalog and Agentic Storefronts | 11 January 2026 | Shopify | Syndicates merchant products to ChatGPT, Microsoft Copilot, Google AI Mode and Gemini |
| Model Context Protocol (MCP) | Open standard | Agentic AI Foundation (Linux Foundation) | Connects AI applications to external systems and data |
The landscape is moving. OpenAI launched Instant Checkout for US Etsy sellers in September 2025. In March 2026 OpenAI said Instant Checkout was moving to retailer apps inside ChatGPT, while ACP development continued. At Google Marketing Live on 20 May 2026, Google announced Universal Cart across Search and Gemini and UCP checkout in ads. Plan for several protocols in parallel.
What do AI shopping agents need from product data?
Feed specifications show what agents consume. OpenAI's product feed specification requires item_id, title, description, url, brand, seller_name, image_url, availability and price, with gtin and mpn optional in its own format and required in its Google-compatible format unless no identifier exists. Flags such as is_eligible_search and is_eligible_checkout control where products appear. Google's UCP implementation requires an active Merchant Center account with eligible products.
A practical readiness checklist:
- Identifiers. A valid GTIN per trade item, plus brand and MPN. See GS1, GTINs and GDSN.
- Comparable attributes. Specifications with units, materials, dimensions and certifications as typed fields.
- Variant structure. Item groups with explicit options for size, colour and configuration.
- Commercial freshness. Price, availability and shipping updated whenever they change.
- Policies. Returns, warranty and delivery terms in machine-readable form.
- Context. Compatibility, use cases and Q&A, which also serve generative engine optimisation.
- Images. Direct image URLs per variant that meet channel requirements.
How does PIM fit into agentic commerce?
A PIM (product information management) system holds the governed product record, while commerce, ERP and order systems hold live price, stock and checkout. Agentic commerce needs both, joined in feeds or APIs. That makes PIM integration architecture a prerequisite: which system publishes each field, how often and through which channel connector.
PIM vendors are adding agent access. Stibo Systems, inriver and Bluestone PIM advertise MCP servers that let AI agents query governed product data, and Akeneo publishes guidance on preparing product data for agentic commerce. For brands selling through retailers, product data syndication remains the main route, as retailers feed the agents.
What should teams do now?
Start with an audit of the top revenue SKUs against the OpenAI and Google feed specifications. Measure completeness of identifiers, core attributes and variant data, then close gaps at the source in PIM. Assign an owner for each agent channel, decide which products are eligible for agent checkout and monitor how products appear in assistant answers.
Key takeaways
- AI shopping agents discover and buy through structured feeds and open protocols, chiefly ACP, UCP, AP2 and MCP.
- Protocols and programmes changed several times between September 2025 and May 2026, so build channel-neutral product data.
- Agents need identifiers, typed attributes, variant structure, current price and availability, and clear policies.
- PIM supplies the governed record, while commerce and ERP systems supply live price and stock.
- Audit top SKUs against published feed specifications and fix gaps at the source.
Frequently asked questions
What is the Agentic Commerce Protocol?
The Agentic Commerce Protocol (ACP) is an open standard co-developed by OpenAI and Stripe, announced on 29 September 2025 with Instant Checkout in ChatGPT. It defines how merchants share product feeds and handle checkout and delegated payment with AI agents. It is published under the Apache 2.0 licence, with the latest stable release dated 17 April 2026.
What is Google's Universal Commerce Protocol?
The Universal Commerce Protocol (UCP) is an open-source standard announced by Google on 11 January 2026, developed with Shopify, Etsy, Wayfair, Target and Walmart. It covers the commerce journey from discovery to order management, works with the Agent Payments Protocol and supports REST, MCP and A2A connections. Google's implementation requires a Merchant Center account.
Do B2B manufacturers need to prepare for agentic commerce?
Yes, in a different form. B2B buyers and procurement systems increasingly use AI assistants to search catalogues and compare specifications. Manufacturers benefit from the same groundwork: verified identifiers, classified products, complete technical attributes and documents, and an API or MCP server through which authorised agents can query governed product data.
For coverage of PXM, PLM and product compliance, including PPWR and DPP, see the independent resource PXM Atlas.