Travel metasearch as we know it today — Google Hotels, Kayak, Skyscanner, and Trivago — is facing a new reality in the agentic AI era.
Travelers will still want a deal when viewing hotel rates and airfares on AI platforms such as ChatGPT, Gemini or Claude. But it’s not clear what form price comparison will take, and how legacy metasearch platforms will fare.
Skift spoke with a current CEO, two former CEOs, and several former C-suite executives from metasearch companies to get their takes on how the future will shake out.
The most likely scenario: AI agents will automatically compare prices using APIs from Google Flights and Skyscanner for airfares; Booking.com, Expedia, and hotel chains for accommodations; and CarTrawler and Travelport for rental cars, for example.
In theory, the results would be personalized for travelers because the AI platforms know their travel histories and preferences. They’d also know everything from whether a traveler qualifies for a senior discount to Delta SkyMiles and Marriott Bonvoy tiers.
That’s what metasearch was supposed to do, but never quite managed.
“Metasearch promised us that it was going to provide comprehensive results, personalized for me, presented in an objective fashion,” said Drew Patterson, venture partner at Tidemark and former CEO of hotel metasearch engine Room 77, which was acquired by Google. “And the reality is today, the results on a on a metasearch engine come from those sources that they get feeds from, presented in a way that are homogeneous. They’re the same for all kinds of customers, and they’re likely influenced by the monetization.”
Because AI platforms are acting as agents of the customer, the personalization and richness of their offers may be far greater than what travelers can view on metasearch engines, Patterson said.
Trivago CEO Johannes Thomas said the future of metasearch will see travelers “comparing offers that come from AI agents and agent-based booking experiences alongside the traditional OTA and hotel-direct rates.”
Thomas said AI platforms will likely start with single partners, but over time will be comparing pricing across multiple sources.
In addition to LLMs making countless API calls to external systems such as Booking.com and Google Flights to check-prices, the LLMs could decide to build in-house price-comparison tools that live on their own servers that would rely on predictions and cached data for speed.
Thomas cautioned it would be easy to underestimate the complexity for AI platforms to construct a great travel shopping and booking experience because it requires managing relationships with thousands of suppliers.
“Even for a company like Google, it took more than a decade to build out travel and hotels in a meaningful way,” Thomas said. “That shows how long it can take to develop the infrastructure, know-how and operational excellence, even if you have deep pockets.”
In addition to agentic features from OTAs and the LLMs themselves, Thomas expects a wave of agent services that suppliers deploy as new distribution channels. “I don’t expect this to happen overnight, but I do see it evolving steadily over time,” Thomas added.
Thomas said Trivago is “talking to a number of LLM and AI players” about powering their hotel verticals. He said Trivago’s Book & Go feature can facilitate bookings for them on behalf of partners.
There are also numerous technical barriers to accessing real-time personalized rates from OTAs, airlines, and accommodation providers. Much of this data can only be accessed via contract, might only be available on a delay, and hotel tech stacks aren’t set up for individual-level pricing.
“If you look at the infrastructure today, it is mostly pull — you send requests to your table of cached data to gain speed as live requests would be too slow,” said a former top executive at a metasearch firm. “For searches of popular places in the next days 90 days, the cache would have the data, but if there is no cached price, a live request is sent.”
That means AI platforms might have to build their own price feeds, over even consider acquiring an existing metasearch player.
“If I were running the travel vertical at ChatGPT, I would just acquire Trivago or Kayak for a few hundred million dollars and hook up the back-end plus the existing commercialization contracts to my system,” the former metasearch official said. “Easy.”
Another issue in providing price comparisons is to match one hotel versus another. “If they assume hotels are similar and they actually aren’t, it will affect the quality of results,” the former executive said.
Martin Lumbye, the CEO of Carama Family Office and the former CEO of Momondo, acquired by Kayak, said travelers will still lean toward trusted brands to ensure they aren’t getting ripped off.
However, the process may not start with a search. Platforms may use AI to send travelers “totally personalized trip proposals based on all they know about you from a whole life of travel or from data created from your profile.”
The content sent to travelers would be inspirational and visual “instead of just waiting for someone” to enter their desired origin and destination, Lumbye said.
Thomas of Trivago said he expects there will be a metasearch interface that would compare rates from AI agents and those from hotels and OTAs.
One source said there would be a solution needed to compare rates of LLM platforms, such as the trip recommended on Gemini versus ChatGPT.
Those rates would need to be personalized to be relevant. One metasearch veteran said such a solution would be coming soon.
Get breaking travel news and exclusive hotel, airline, and tourism research and insights at Skift.com.
ThePostMaster
