AI agents and the end of the search box

It is now plausible — not certain, but plausible — that a meaningful share of travel intent in 2027 will be captured before the traveler ever reaches an OTA, metasearch, or supplier site. AI assistants are becoming the place where the trip is conceived, even if they are not yet the place where it is booked.

This is the strategic fact that travel distribution needs to internalize. Everything else is downstream of it.

What's actually happening

A small but growing share of leisure trip planning is starting in conversation with ChatGPT, Claude, Gemini, and Perplexity. Skift reported in March 2025 that more than 30% of US leisure travelers under 35 had "consulted an AI assistant for trip planning at least once" in the prior six months — a number that was effectively zero in 2023.

The bookings themselves still happen on traditional surfaces. But the intent — the destination decision, the property shortlist, the rough dates — is increasingly formed somewhere upstream.

This is a different problem than "users using AI." It is a question of where the high-value moment in the funnel migrates to.

The three positions in the new stack

We see three viable strategic positions emerging.

• Be the agent. Build or partner on the conversational layer itself. Realistic only for a small number of players: the hyperscalers, the largest OTAs (Booking's OpenAI partnership announced in late 2024 is the proof point), and a handful of well-funded startups like Mindtrip and Layla. The economics are uncertain but the strategic position is the strongest.
• Be the inventory layer the agents prefer. Win on data quality, API responsiveness, deep linking, and commercial terms that make a given supplier or aggregator the natural answer when an AI assistant picks one. Less glamorous, very defensible. This is where most large OTAs and channel managers should be playing.
• Be the brand the user asks for by name. When a user says "find me a Marriott Bonvoy property," the agent obliges. Brand equity becomes a defense against disintermediation. This is the under-discussed leverage for the largest hotel and airline loyalty programs.

The trap is to do none of these well — to treat AI as a UX widget while the structural game is happening at the intent layer.

What investors should watch

Two leading indicators will tell you who is winning.

The first is API call volume from large language model providers to supplier and aggregator APIs. This is largely invisible publicly but very visible to suppliers themselves. If you can see it, you have an information edge. The Information reported in 2025 that Trip.com's internal dashboards show "agent-mediated" supplier calls growing at triple-digit rates quarter-on-quarter.

The second is the share of bookings on a given supplier that arrive without a traditional referrer — neither paid search, nor direct, nor organic in the classical sense. As this category grows, it is a proxy for AI-mediated traffic.

The read-through

For OTAs: the existential question is whether you become the agent or the inventory. Both are defensible. Splitting the difference is not. Booking is going for both ends; Expedia has been clearer about playing the inventory game; Trip.com has been the most aggressive in agentic UX.

For hotels and airlines: this is the strongest argument in years for direct-channel investment, but the investment should be in being agent-ready — clean APIs, structured content, transparent rates — rather than in another redesign of the homepage. Bloomberg reported in mid-2025 that Marriott and Hilton both have internal "agent-ready" programs running.

For metasearch: the position is uncomfortable. Metasearch's value proposition is comparison, which agents do natively. Survival means becoming a wholesale data layer agents depend on (the Kayak/Booking strategy), or evolving the consumer brand into an agent itself (the path Tripadvisor seems to be choosing).

For investors: the next 18 months are a land grab. The winners will look obvious in hindsight and not obvious now. That is exactly when the alpha is.

The search box is not disappearing tomorrow. But pretending the question of where intent lives is settled is the most expensive mistake a distribution leader can make right now.