Agentic commerce refers to online trade in which AI agents independently research, compare products and prepare or execute purchase decisions on behalf of people. Instead of a human clicking through a shop, an agent queries structured product data. A shop only takes part if it is machine-readable: clean product schemas, current feeds and agent-ready interfaces such as MCP.
An AI agent — for example in ChatGPT or an assistant — is given a task like a product search with a budget. It queries structured data and interfaces, compares options and presents the user with a selection, sometimes up to checkout.
The basis is machine-readable, reliable data — not pretty HTML. What an agent can't cleanly read effectively doesn't exist for it.
Complete Product and Offer schemas with price, availability and variants. Current product feeds. Clearly named entities. Ideally an agent-ready interface such as an MCP server through which an agent queries catalog and stock directly.
It's the same foundation as for GEO: whoever is visible for AI answers is also connectable for buying agents.
The shift is early but real: payment providers and AI vendors are building standards for agent checkout. Whoever makes their shop agent-ready early is visible when buying agents go mainstream.
The effort is manageable because much of it builds on clean structured data you need for SEO and GEO anyway.
No. A shop chatbot helps the human on the page. Agentic commerce means an agent outside your shop queries your data and prepares the purchase for the user.
For humans yes, for agents often not. Agents need structured, reliable data — otherwise they guess from HTML or skip the shop.
The Model Context Protocol is a standardized interface through which an agent retrieves catalog, prices and stock directly and reliably, instead of scraping a website.