Guides
The Year AI Started Shopping: Agentic Commerce and Magento
October 15, 2025 - by Alexandru-Manuel Carabus
In 2025 shoppers started buying inside AI chats, not just on websites. Here is what agentic commerce is, the traffic shift behind it, and where Magento fits.
For twenty years, online shopping meant search, then click, then buy on a website. In 2025 a second path opened: ask an AI assistant, then buy inside the chat. ChatGPT began completing purchases directly, and the commerce world had to take agentic commerce seriously.
What agentic commerce actually is
Agentic commerce is the ability to discover and buy a product inside an AI assistant, without ever landing on the store. The shopper asks the assistant to research, compare and order, and the order completes in the conversation. Two emerging standards, the Agentic Commerce Protocol and the Universal Commerce Protocol, define how a store connects to that flow, and Adobe committed to supporting them.
The traffic shift is real
This is not a thought experiment. Adobe Analytics reported that generative AI referral traffic to United States retail sites was up roughly 1,200 percent year over year in early 2025, and around 4,700 percent by mid-year. Shoppers arriving from AI referrals converted about 31 percent higher and generated far more revenue per visit than traditional traffic. The audience moved first, as it usually does.
Where Magento stands
Because Magento is open, the connection to agentic protocols does not depend on a vendor shipping a feature. Community open-source modules already bridge Magento to the Agentic Commerce Protocol for ChatGPT checkout, and Adobe committed to the standards at the platform level. An open stack gets to adopt this on its own timeline rather than waiting in a roadmap queue.
Agentic commerce is not the same as AEO
It is worth separating two things people blur together. Answer engine optimization is about being the store an AI recommends. Agentic commerce is about being the store an AI can actually buy from. You want both, and they need different work: clean structured data and reputation for the first, a working product feed and a supported checkout protocol for the second.
What to do now
The near-term work is unglamorous and useful anyway: accurate structured data, a clean and complete product feed, and catalog hygiene, so that when an assistant reads your catalog it gets the right answer. Then watch the Agentic Commerce Protocol and decide when to connect. If you want to map this to your own store, book a free strategy call with LIQUIDLAB.