Trust' became online travel's favorite word this week — and two camps meant opposite things by it. The incumbent giants are betting travelers keep returning to known brands because a general-purpose model might hallucinate a booking; a larger rival is betting the real contest is earning the trust of the AI agents themselves, exposing verified inventory and live pricing as machine-readable rails so demand routes through it whether or not anyone opens an app. That split is the actual story, and it rhymes with everything else on the wire: a payment network moving upstream from the swipe into curated discovery, a global leisure platform walking into managed corporate travel just as the domestic incumbent that owned that lane slides toward a delisting notice, a state rail monopoly absorbing premium-demand growth through a channel it controls and is mid-rebuild on. Even a booking assistant's personality — how it hedges, whether it asks before acting — turns out to be an unowned trust surface no one deliberately drew. The question for the category: when discovery, settlement, corporate supply, and the agent's own voice each belong to someone else, what is the aggregator still the owner of?
Assistants built on the same model and instructions still feel like different people — one hedges and defers, another acts then reports — and the argument is that these traits emerge as an accidental byproduct of alignment, not deliberate design. The proposal: treat AI personality as an interface layer chosen the way a team chooses a button state, rather than letting it arrive undrawn.