Google announced the Universal Commerce Protocol from the main stage at NRF, then wrote it up in the slightly bureaucratic register its blog reserves for its own biggest news. Read the post twice and the substance is clear enough — and significant.
UCP, co-developed with Shopify and a group of large retailers including Etsy, Wayfair, Target and Walmart, is designed to plug brands directly into AI shopping conversations across Search, Gemini and AI Mode. Three pieces of plumbing stand out. Business Agent lets a shopper chat with a brand on Google Search the way they’d chat with a virtual sales associate — questions, comparisons, recommendations, all inside the conversation. Direct Offers surfaces discounts at the moment of stated intent rather than the moment of bored scrolling. And a new set of Merchant Center attributes feeds the conversational layer with the structured signal it needs to answer confidently.
That’s Google telling its merchants, in writing, that on its own surfaces the shopping journey happens inside conversations rather than ten blue links.
The protocols are converging
The bigger story is convergence. OpenAI has been wiring checkout into ChatGPT. Shopify has put agent-facing storefronts across ChatGPT, Copilot, Gemini and Google AI Mode. Walmart pulled back from third-party agentic checkout — we covered that separately — but the platform-side trend hasn’t slowed; it’s accelerated. Now Google is in the same shape, and at pains to position UCP as a protocol the broader ecosystem can implement, not a Google-only standard.
The pattern is unmistakable. Whether your stack runs on Shopify, WooCommerce or something bespoke, the AI assistant is now a discovery surface — and increasingly a transaction-adjacent one.
The tighter question
Talk to founders in supplements, beauty, kit and accessories about this news and the reaction follows the same arc in every conversation: first interest, then quiet, then a tighter question.
The tighter question is always some version of: how do I show up?
The honest answer is uncomfortable for most catalogues, because it doesn’t reward the part of the budget they’ve invested most heavily in. You don’t show up in AI answers by having a louder paid social presence. You don’t show up by paying more for retargeting. You don’t show up because your homepage hero is well-designed.
You show up because the assistant — a probabilistic, language-driven model with a recommendation surface — has enough coherent, structured information about your product to summarise it in a paragraph of natural language without hedging. Ingredients explicit. Claims sourced. Comparisons present. The same language across the site, the spec sheet, the FAQ and any third-party listings that carry authority. Schema that doesn’t contradict the visible page.
Most catalogues haven’t done that work. They’ve done a version of it for Google’s old crawler, optimised against a query-and-page-rank model that the assistant has, on the surfaces where shopping now starts, stopped using.
The window
Predictions in this category age badly — the release cadence has outrun everyone’s forecasts. But the directional read is hard to escape: within a year, the question for any consumer brand of meaningful size will not be are we doing AI discovery but are we doing it well enough to be the brand the assistant recommends rather than the brand it mentions in passing. The first wave of winners will be the brands that started restructuring their product data this year, not the ones waiting for a board mandate next year.
If you haven’t opened ChatGPT or Gemini and asked it to recommend a product in your category this week, do it before Friday. The answer is the thing your customers are about to hear.
Then get the systematic version: your free GEO Score.