AI shopping bots mostly see airline fares through booking networks and travel sites — so who controls that pipe gains quiet pricing power?

This explores whether AI shopping agents shift pricing power from airlines to the systems in between: the booking networks, travel sites and agent platforms that decide which fares an agent sees. The short answer from the corpus is that power is moving to these middle layers, but the evidence shows this as a new pressure point, not a finished transfer. The clearest example is Amadeus, one of the big systems that sends airline fares out to travel sellers. AI agents now run up to 200,000 searches for each ticket sold, and pricing every one of those searches live became too expensive. So Amadeus precomputes fares for common routes and dates and stores them in a cache Does caching fares give infrastructure providers pricing power?. The surprising part is that power hides in a technical setting. Whoever decides how often the cache refreshes, and which routes get cached, decides which offers an agent can see. An airline can set any price it wants, but an agent may never see it if the cache shows an older one.

Travel more broadly suggests intermediaries get stronger, not weaker. Many predicted that AI agents would cut out online travel agencies. Instead, independent hotels came to rely on those agencies more, reaching 63.4% of bookings. The agencies kept their role as the official seller of record through API deals, and Expedia's business-to-business segment grew 24% Are AI agents actually removing online travel agencies from bookings?. Prieto explains the split: agents make searching cheap, but payment, supply contracts, trust and completing the transaction stay with the platforms Are AI agents actually disintermediating online travel platforms?. Applied to airlines, the bargaining power sits wherever agents depend on someone to fill the order and handle the money, not wherever the search happens.

There's a counterweight to watch. Choudary argues that agents acting only for buyers change where shoppers enter a market, not how the market works. Real change would require airlines to field their own agents that negotiate back Do one-sided consumer agents actually change market structure?. Formal models point to a darker possibility. When a platform charges for each look at a seller, it can push buyers to over-search. Consumers end up no better or worse off, while value moves from sellers to the platform Does platform pricing of inspection tokens encourage excess search?. Swap "sellers" for "airlines" and you have a mechanism by which agent search volume itself drains margin toward the infrastructure.

Two more threads complicate the picture. First, if airlines answer with their own pricing agents, the UK competition authority (the CMA) warns that such agents can drift into quiet coordination without any explicit agreement Could autonomous pricing agents dampen competition without explicit collusion?. Experiments find that pairing a large model with a smaller one makes this coordination more stable, because the larger one leads and the smaller one follows Does model-size difference break LLM pricing collusion?. That pattern resembles a dominant carrier and its smaller rivals. Second, whether intermediaries end up holding the power depends on how the rules are set. Shared protocols can open markets to many competitors Do protocols reshape markets by enabling multi-vendor competition?, and data portability keeps switching possible Should data portability be infrastructure for agent-driven markets?. Without them, a few platforms could own the layer that connects agents and airlines Will agent economies develop as open webs or closed gardens?, Is Google building a chokepoint across data, chatbots, and payments?.

A gap worth naming: the corpus has one airline-specific source, the Amadeus caching note, and no direct data on airline margins or fare outcomes. The best-supported takeaway is the one you might not expect. In agent-driven travel, how fresh the cached fares are may become as important as the price itself, and whoever runs the cache holds that lever.