Amid the recent rapid evolution of search and new shopping methods in the age of agentic and generative AI, Google introduced new protocols and tools for retailers and advertisers.
On Jan. 11 at the National Retail Federation show in New York, the AI technology giant revealed that it has partnered with e-commerce platforms such as Shopify, Walmart, and Etsy to develop the Universal Commerce Protocol (UCP). The open standard supports the shopping experience using AI agents.
The new protocol emerges amid a shift in online shopping, in which consumers are increasingly turning to AI search engines such as Gemini, GPT from OpenAI, and Perplexity to find what they need, rather than visiting retailers directly.
“The reality is that consumers are already leveraging AI to discover products,” said Liz Miller, an analyst at Constellation Research. “Businesses will need to follow customer trust if they don’t want to be left behind.”
Not only are consumers using AI tools, but AI tools are also becoming an increasingly integral part of the search experience. This means that merchants and advertisers must find a way to appear within the search engine themselves, so their businesses are not negatively affected.
Universal Commerce Protocol
The protocol presents what Google already has in its listings and places it in the vendor’s AI Mode or Gemini model, while providing autonomy to AI agents to make purchases from those listings, said William McKeon-White, an analyst at Forrester.
“This protocol seems exclusively focused on helping agents interact with retailers,” he said. “This looks like it’s just giving agents the tools they need to shop accurately and access the right integrations.”
He added that it will be interesting to see how much independence agents will get to perform tasks such as finding lower-priced options or comparing retailers.
One feature of UCP is Direct AI Checkout within Google AI Mode and the Gemini app. This allows users to make purchases without being directed to a retailer’s website.
According to Google, UCP works in conjunction with other generative AI standards, such as Anthropic’s Model Context Protocol and Google’s Agent Payments Protocol (AP2). Google launched AP2 last September; it allows AI agents to make purchases on behalf of both consumers and merchants. With AP2, Google is working with payment vendors like Mastercard and PayPal, while with UCP, it works with e-commerce websites and users can make purchases from the sites without leaving AI Mode or Gemini.
“The promise of the UCP standard is that it offers any merchant a means to be discovered on a Google surface,” Miller said.
Google is not the only vendor to provide such promises. Last September, OpenAI released its own Agentic Commerce Protocol, in partnership with e-commerce payments vendor Stripe. The OpenAI protocol also enables buyers to make purchases in ChatGPT from participating merchants.
Merchants and Advertisers
Even though many AI vendors are offering diverse ways for merchants to be discovered within generative AI systems, Google has the edge due to its experience in building a commerce backbone, Miller said.
For merchants, the challenge is ensuring they diversify and are not relying solely on protocols from Google or OpenAI, she said.
“They will also need to ensure that content on their site can be discovered by the LLMs themselves, meaning businesses and brands will need an answer engine optimization and generative answer optimization strategy in place,” Miller added. These approaches create content based on questions that might be asked in an AI search engine.
Other Tools
Aside from the UCP, Google introduced other tools. One is Business Agent, which lets shoppers chat with brands in Search. The agent is live with retailers such as Lowe’s, Michaels, Poshmark, and Reebok.
Another tool is Direct Offers. It enables advertisers to present exclusive deals to shoppers who are ready to buy a product in AI Mode.
The new tools present benefits but also risks for Google, advertisers and merchants, Miller said.
“The merchant has to weigh if they can afford to stay out of this model, similar to how merchants and advertisers need to weigh if they can afford not to be in the sponsored results,” she said.
Google also faces a perhaps even greater risk if a product purchased in AI Mode does not meet expectations, it could erode trust in the LLM.” Consumers have already said, time and again, that they will walk away from an agentic/machine-driven solution after one failure,” Miller said. “What they are really saying is there is more room for empathy or forgiveness when a human is involved.”
This means that there must be an effective mix of agent behavior, data accuracy, and the quality of the shopping experience, McKeon-White said.
“If too much agent behavior is enabled, it may result in unexpected outcomes, excess deals, or unexpected and incorrect product purchasing,” he said. “If the listing information is incorrect, that impacts a purchase faster. If the experience is bad or degraded from over-suggestion or too many ads, it may be that customers abandon the experience Google is designing.”
