Shortly after Google announced its new Universal Commerce Protocol for AI-powered shopping agents, a consumer economics watchdog sounded the alarm.

In a now viral post on X viewed nearly 400,000 times, Lindsay Owens on Sunday wrote, “Big/bad news for consumers. Google is out today with an announcement of how they plan to integrate shopping into their AI offerings including search and Gemini. The plan includes ‘personalized upselling.’ I.e. Analyzing your chat data and using it to overcharge you.”

Owens is executive director of the consumer economics think tank Groundwork Collaborative. Her concern stems from looking at Google’s roadmap, as well as delving into some of its detailed specification docs. The roadmap includes a feature that will support “upselling,” which could help merchants promote more expensive items to AI shopping agents.

She also called out Google’s plans to adjust prices for programs like new-member discounts or loyalty-based pricing, which Google CEO Sundar Pichai described when he announced the new protocol at the National Retail Federation conference.

After TechCrunch inquired about Owens’ allegations, Google both publicly responded on X and spoke with TechCrunch directly to reject the validity of her concerns.

In a post on X, Google responded that, “These claims around pricing are inaccurate. We strictly prohibit merchants from showing prices on Google that are higher than what is reflected on their site, period. 1/ The term “upselling” is not about overcharging. It’s a standard way for retailers to show additional premium product options that people might be interested in. The choice is always with the user on what to buy. 2/ “Direct Offers” is a pilot that enables merchants to offer a *lower* priced deal or add extra services like free shipping — it cannot be used to raise prices.”

In a separate conversation with TechCrunch, a Google spokesperson said that Google’s Business Agent does not have functionality that would allow it to change a retailer’s pricing based on individual data.

Owens also pointed out that Google’s technical documents on handling a shopper’s identity say that: “The scope complexity should be hidden in the consent screen shown to the user.”

The Google spokesperson told TechCrunch that this is not about hiding what the user is agreeing to, but consolidating actions (get, create, update, delete, cancel, complete) instead making a user agree to each one separately.

Even if Owens’ concerns about this particular protocol are a nothingburger as Google asserts, her general premise is still worth some thought.

She is warning that shopping agents built by Big Tech could one day allow merchants to customize pricing based on what they think you are willing to pay after analyzing your AI chats and shopping patterns. This is instead of charging the same price to everyone. She calls it “surveillance pricing.”

Although Google says its agents can’t do such a thing now, it’s also true that Google is, at its heart, an advertising company serving brands and merchants. Last year, a federal court ordered Google to change a number of search business practices after ruling the company was engaged in anticompetitive behavior.

While many of us are excited to welcome a world where we’d have a team of AI agents handling pesky tasks for us (rescheduling doctor’s appointments, researching replacement mini-blinds), it doesn’t take a clairvoyant to see the kinds of abuse that will be possible.

The problem is that the big tech companies that are in the best position to build agentic shopping tools also have the most mixed incentives. Their business rests on serving the sellers and harvesting data on consumers.

That means AI-powered shopping could be a big opportunity for startups building independent tech. We’re seeing the first few sprinkles of AI-powered possibilities. Startups like Dupe, which uses natural language queries to help people find affordable furniture, and Beni, which uses images and text for thrifting fashion, are early entrants in this space.

Until then, the old adage probably holds true: buyer beware.