Could AI agents that haggle and compare on our behalf actually give consumers more market power?

This explores whether there's real evidence that communication costs (the effort of searching, asking, comparing and bargaining) currently narrow consumer choices. The short answer is that the corpus has a strong argument for this idea and very little direct measurement of it. The clearest statement comes from Microsoft researchers, who argue that agentic AI matters economically less because it automates tasks and more because it lowers the friction of consumers and businesses talking to each other. In their view, that could reduce switching costs and shift market power toward consumers Could agent-to-agent communication reshape market power?. The note is candid that this support is conceptual rather than empirical. It describes a hypothesis, not a finding.

The closest thing to evidence comes from formal economic models rather than field data. One model shows that consumers rationally keep searching by hand even after agentic search would raise platform revenue. Platforms judge agents only by conversions, while consumers also pay the cost of the search effort itself Why do consumers delay adopting agentic search?. That gap is an indirect sign that search and communication costs shape behavior. Another model shows that when a platform charges for 'inspection tokens' (the right to look closely at a seller), its profit-maximizing price pushes buyers to over-inspect. Surplus moves from sellers to the platform, and consumers end up no better off Does platform pricing of inspection tokens encourage excess search?. Both treat communication cost as a real lever on choice, but both are theory.

The more surprising thread is that lowering communication costs doesn't automatically widen choice. A 2026 market design paper argues that nearly free agent search, combined with sellers generating near-identical descriptions and low-quality entrants flooding in, can leave markets worse off than before Can cheap agent search actually harm shopping markets?. And the agents doing the communicating bring their own distortions. In head-to-head tests, LLM shoppers swing far harder than humans in response to price, rating and nudge cues Do shopping agents respond more strongly to price and rating cues?. They also favor sponsored products even at double the price Do language models favor expensive sponsored products over cheaper alternatives?. In negotiations, buyer agents reveal their budgets unprompted and overpay Do LLM negotiation agents reliably follow budget and pricing rules?.

One finding flips the usual framing. Cheaper communication can leak more than it gains. When a buyer speaks through a natural-sounding shopping persona, a seller-side agent can work out their real willingness to pay almost perfectly from the conversation alone. A confidential numeric budget leaked far less Can a seller infer what a buyer really wants to pay?. So the friction that limits consumers may also have been protecting them.

To be direct about the gap: nothing in this set measures how communication costs restrict real consumer choices today, such as how many options people never consider, or how often switching fails because it takes too much effort. Survey data shows consumers expect agents to reshape their spending Will brands succeed by building their own AI shopping agents?, but expectations aren't evidence of a current constraint. The idea that friction limits choice is widely assumed here and modeled carefully, but it hasn't been demonstrated empirically.