Travel-booking AI agents are great at search — but why haven't they knocked out Expedia, Booking, or Airbnb's stock price?

This explores why AI disruption hits online travel businesses unevenly. The corpus doesn't have a side-by-side breakdown of OTA categories such as hotels, flights, rentals, and metasearch. What it does have points to one answer: the unevenness depends less on the travel category than on which part of the booking chain a business owns. A booking has two halves. Discovery is searching, comparing, and recommending. Fulfillment is payment, supplier contracts, customer trust, and completing the transaction. AI agents are very good at the first half and mostly need someone else to handle the second.

The clearest statement of this is in Are AI agents actually disintermediating online travel platforms?. Expedia, Booking, and Airbnb all posted strong 2026 gains despite years of warnings that agents would cut them out. Prieto's explanation is that AI makes search cheap but leaves untouched the functions these companies actually control: settling payments, holding supply, and absorbing risk when a trip goes wrong. Value moves around rather than disappearing. A business whose main asset was helping you find things is exposed. A business whose main asset is completing the booking is not.

Are AI agents actually removing online travel agencies from bookings? makes this sharper and shows where the gaps open up. Goswami finds OTAs being built *deeper* into the system. They act as the merchant-of-record (the party that legally takes the payment) behind AI agents through API connections. Independent hotels became *more* dependent on OTAs, reaching 63.4% of bookings, because they lack the infrastructure to plug into agents themselves. Expedia's business-to-business segment, which sells booking plumbing to other companies, grew 24% and now makes up 38% of revenue. So one source of unevenness is business model: OTAs that already sell infrastructure gain from AI agents, while those that depend on consumers arriving at their own front page have more to lose.

For the downside, look outside travel. Do AI Overviews actually reduce clicks to publisher sites? found in a randomized experiment that Google's AI Overviews cut clicks to publisher sites by almost 40%, with no gain in user satisfaction. That is what happens to a business that only does discovery: the AI answers the question and the visit never happens. Travel players that look like publishers, such as review sites, comparison pages, and metasearch engines that live on referral traffic, are structurally closer to this case than to Expedia's.

The takeaway you might not expect: AI agents may be *increasing* concentration in travel. Suppliers like independent hotels can't easily connect to agents directly, so they lean harder on the big intermediaries. The likely losers are the thin discovery layers that sit above the transaction, not the OTAs themselves. If you want category-level evidence, such as airlines versus hotels or package tours versus short-term rentals, the corpus doesn't cover it yet.