Ulta Is Betting Open Standards Can Beat Amazon at AI Shopping

When Ulta Beauty and Google unveiled their Gemini-powered shopping partnership this week, the press release read like every other AI retail announcement: personalized recommendations, seamless checkout, the future of beauty discovery. Flip past the quotes from chief merchandising officers and you find something more durable underneath. The deal runs on something called the Universal Commerce Protocol, a technical standard that its architects — Google and Shopify — are positioning as the connective tissue for a new kind of shopping.

UCP is an open standard: any retailer, any platform, any AI agent can build against it without licensing a proprietary system. Twenty-plus retailers and platforms have endorsed it so far, according to Shopify's documentation. That is what makes this week's announcement the first live test of whether the open-standard model can produce results comparable to what a single company controlling the full stack can deliver on its own.

The pressure point is Walmart. One month before Ulta announced its Gemini integration, Walmart ended its partnership with OpenAI's Instant Checkout feature. Conversion rates ran roughly three times lower than on Walmart.com directly, The Street reported, citing accuracy and integration problems. Walmart replaced the OpenAI checkout with its own agent, Sparky, embedded in ChatGPT and Gemini instead of letting OpenAI handle the transaction end-to-end. The structural lesson from that failure is what the UCP coalition is built to address: UCP is designed to keep the retailer as the merchant of record, with orders flowing into their own systems rather than being intermediated by the platform.

The Amazon comparison is the one the industry watches most closely. Amazon's Rufus agent drove a 90 percent increase in shopping sessions from October through Black Friday last year, according to Marketing Week, while non-Rufus sessions on Amazon grew just 8 percent. Rufus works because Amazon controls the product catalog, the checkout, and the AI. It is vertical integration executing as designed. The UCP model is the explicit alternative: keep the merchant in the middle, let the AI handle discovery and checkout facilitation, and see if the results hold.

The consumer trust question is the other load-bearing variable. Sixty-four percent of shoppers say they need at least one safeguard — a money-back guarantee, a confirmation step — before letting an AI agent complete a purchase on their behalf, according to Marketing Week, citing a Harvard Business Review study. Beauty is an intimate category: skin care routines, color matching, fragrance preferences. Convincing someone to let an algorithm buy their moisturizer is a different challenge than convincing them to let it reorder batteries.

Ulta is deploying to 46 million loyalty members, according to the companies' joint announcement, giving the partnership the largest real-world test bed for consumer comfort with AI purchasing that the open-standard model has seen. No conversion data exists yet — the rollout happens over the next month. That absence is the honest version of this story: the infrastructure is assembled, the retailers are lined up, but nobody has yet demonstrated that an open protocol can convert at rates that justify the build.

What happens next is fairly straightforward to sketch. If Ulta's numbers look like Walmart's, the UCP coalition takes a credibility hit right as Amazon's Rufus extends its lead. If the numbers hold, Google and Shopify have a proof-of-concept for the interoperability play that Amazon cannot match without abandoning its own commerce infrastructure — a tradeoff Amazon shows no sign of making.