Google promises its AI pricing tool will only cut prices, never raise them — but what stops a store from hiking the 'list' price first?

This looks at why Google's personalized pricing tool, which lets major retailers show different shoppers different prices through its AI systems, seems to fall outside the rules we already have, and where those rules run out. The corpus points to three separate gaps. The first is about what a 'discount' means, the second is about competition law, and the third is about how consumer law gets enforced at all.

Start with the discount problem. Google's main reassurance is that the tool will only lower prices, never raise them. Can Google's pricing tool really prevent price increases? explains why that promise is hard to check. Nothing controls the starting price that the discount comes off, so a merchant can raise the list price and then 'discount' back down to where it always meant to charge. Stoller compares this to hospital chargemasters and pharmacy markups. Is Google's personalized pricing actually surveillance pricing? goes further and calls the whole thing surveillance pricing: personal data is used to work out the most each person will pay, and the result is presented as a deal. Airlines and medical billing have done this for years. The gap is that rules aimed at price increases never trigger when every price on screen is labeled a discount. Both of these are arguments, though. Neither note shows direct evidence that consumers are actually overpaying.

The second gap is the less obvious one, and it's in antitrust. Does Google's AI pricing tool coordinate prices across rival merchants? notes that when rival merchants all hand their pricing to the same Gemini service, Google stops being a vendor serving competing clients and starts acting as a middleman that sets prices across them. Price-fixing law is built around finding an agreement between competitors, and here there may be none. The UK's competition regulator, the CMA, describes the same risk more broadly in Could autonomous pricing agents dampen competition without explicit collusion?: pricing agents can learn to soften competition simply by adapting to each other, with no explicit deal. Experiments in Does agent heterogeneity prevent LLM pricing collusion? make this concrete. Two identical AI pricing agents kept prices 22% above the competitive level, and the collusion fell apart once the agents ran on different algorithms. A single shared Google model is close to the worst case, because it makes every merchant's pricing agent identical.

The third gap is structural. Is Google building a chokepoint across data, chatbots, and payments? places pricing inside a larger buildout: personal data, chatbot distribution, retailer deals and payment partnerships all joined at one company. Can EU consumer law rules survive agent-to-agent API transactions? explains why consumer law struggles with this. EU protections assume a person looking at a screen, with required buttons and disclosures. When agents transact with each other through APIs, there is no screen for those rules to apply to. Will sellers game Google's new agentic commerce protocol? adds that sellers have always had reasons to hide their real prices from machines, whether web crawlers, search engines or now agents.

The proposals in the corpus don't close this gap yet. Could federal AI agent registries solve platform access unfairness? would make platforms admit outside agents and give agents a duty to act in the user's best interest. That deals with agents being blocked, not with a platform setting prices for rival merchants. Here's the part you might not have expected to want to know: the most promising fix may not be a disclosure rule at all. Since the experiments show that identical agents collude and diverse ones don't, the stronger protection could be making sure merchants don't all run the same pricing model.