The five kinds of vertical travel AI
Supply, corporate, curation, destination, classification — five kinds of specialist, five different defenses
The travel industry is betting that the way to beat Gemini and ChatGPT is to build a specialist travel AI. The candidates everyone names are Airbnb and the major chains. They are one kind of specialist. There are at least four others, and they defend on completely different ground.
A Navan AI is a vertical travel specialist. So is a Virtuoso AI. So is a Switzerland Tourism AI. So is a sustainability classification AI. None of them competes the way Airbnb does, and the test that works for Airbnb does not work for any of them.
Five types are worth distinguishing.
Supply-owning specialist
The kind that gets most of the attention. Airbnb, Marriott, Hilton, IHG, Hyatt, Accor. Booking and Expedia in their own lane. These specialists own the booking, or sit close enough to it to act.
The defense rests on three things. Execution, because the specialist is connected to the systems that actually run the stay — PMS, loyalty record, the authority to waive a fee or upgrade a room. A horizontal can suggest a room change at eleven at night. The Marriott agent can make it happen. Inventory, because the specialist can guarantee availability, hold a rate, and offer terms that exist nowhere else. Relationship, because the loyalty record carries years of preference and history tied to a known traveler. That record is not something a smarter model can replicate. It is a database the horizontal cannot access.
The defense works when all three are present. Airbnb has execution and inventory but a thinner relationship layer than the chains. The major chains have all three within their owned and franchised properties. Booking and Expedia have relationship data and partial execution authority, but the room itself is controlled by the hotel.
This is the type the industry is mostly watching. It is not the only one.
Corporate and managed travel specialist
Amex GBT, BCD Travel, Navan, Egencia. These specialists serve two stakeholders at once — the traveler who books and the employer who pays — and have to satisfy both.
A corporate AI does not win on travel knowledge alone. It has to obey a policy Gemini cannot see, integrate with an expense system Gemini cannot reach, track a traveler the employer is responsible for, and surface negotiated rates the corporation has on file. The selling point against Gemini is not preference matching. It is policy compliance, expense integration, duty of care, and rates that exist only inside the corporate program.
The defenses are policy authority, where the specialist enforces the rules the horizontal does not see. Expense integration, where the booking flows directly into the systems that pay for it. Duty of care, where the specialist knows where the traveler is and can act when something goes wrong abroad. And negotiated supply, where corporate rates and waivers exist that no consumer agent can reach.
Sitting underneath all of this is the GDS. Amadeus, Sabre and Travelport are not corporate AI specialists themselves, but the corporate AI cannot work without them. The negotiated corporate rate sits in the GDS. The hotel content comes from the GDS. The fulfillment runs through the GDS. In an agent-driven future, the GDS becomes the API layer the corporate AI calls — invisible to the traveler, essential to the booking. Whether the GDSs stay as infrastructure or move up the stack into a specialist position of their own is an open question. The technical assets are there. The dominant business model still points the other way.
Gemini can plan a trip. It cannot enforce a corporate policy, push the receipt into Concur, or call a duty-of-care provider when a traveler needs help in another country. The corporate traveler is also a tracked employee, and the AI that serves them has to behave as if it knows that.
Curation and network specialist
Virtuoso is the cleanest example. American Express Travel is another. The larger luxury travel networks fit here too. These specialists do not own hotels. They organize a curated set of them around a segment.
A Virtuoso AI does not need to own the hotels to be defensible. It needs to know which suppliers are vetted, which amenities the network has negotiated, which advisor matches which traveler, and what the segment expects. The selling point against Gemini is not "I have a better model." It is "I will not waste your time on the wrong supplier, I have leverage you do not have on your own, and the trust of the network is doing work a generic recommendation cannot do."
The defenses are different here. Curated access, where the specialist has filtered the supply down to a vetted subset that fits the segment. Network leverage, where the consortium has negotiated commercial terms — upgrades, breakfast, resort credits, late checkout — that no horizontal can reproduce, because the horizontal does not have the contracts. And brand trust within the segment, where the badge signals that the network has already done the vetting.
Gemini will get better at recommending luxury hotels. It will not get the Virtuoso amenities. It will not have the advisor relationship. It will not be trusted the same way by a traveler who has been booking through the network for fifteen years. The moat is real, just not the same shape as the chain moat.
Destination specialist
A category almost empty of credible attempts. Switzerland Tourism, Visit California, Tourism Australia, and the larger national and regional tourism organizations all have the foundations for one. The defense is destination authority, verified local information, and a curated supply network on the ground.
A Switzerland Tourism AI knows which mountain railway runs in shoulder season, which canton has the festival next weekend, what the Swiss Travel Pass actually covers, and which mountain restaurants are still serving the week before the lifts close. Gemini will be approximately right on Switzerland and frequently wrong in the details. The DMO can be authoritative on the details, because it has the working relationships with the cantonal boards, the hotels, and the transportation operators.
The defenses are verified information, which the horizontal cannot match without an explicit data partnership; local supply, where the DMO knows the operators on the ground in ways an aggregator listing does not capture; and integrated infrastructure, where the destination has direct booking through national travel centers and can negotiate packages no aggregator can offer.
Execution authority is partial — DMOs do not own the hotels. But many have national booking infrastructure, payment processing, and direct relationships with operators that come closer to a complete booking than they get credit for. The constraint is institutional, not technical. Most DMOs are not built to ship a credible AI product. The few that decide to will have a defensible position by default.
Criteria and standards specialist
The least visible category. Sustainability classification systems, wellness certifications, accessibility booking layers, kosher and halal networks. These specialists are organized around a criterion — not a category, not a segment, not a place.
They probably do not face travelers directly. The horizontal calls on them when a traveler asks for sustainable hotels and Gemini needs a reliable source for what "sustainable" actually means. The business model is closer to infrastructure than to consumer product. The moat is the verification method, the certified data, and the trust the badge carries.
The defense is not consumer-facing brand. It is method and data. A horizontal can claim to know which hotels are sustainable. The specialist can prove it, because the specialist runs the audit, holds the criteria, and has the certification record. Plug a sustainability classifier into Gemini and the horizontal looks better. Try to replace it and Gemini looks worse, because the methodology underneath is not something a model can generate.
This is where lifestyle hotels, eco-properties, wellness destinations and accessibility-focused operators benefit from a category they do not have to build themselves. Someone else builds the criteria layer. The hotel needs to be findable inside it.
What the typology means
The five types are not really alternatives. They work together — a horizontal calling a destination AI for the local detail, or a curation network filtering its picks with a sustainability classifier. The industry treats horizontal versus vertical as winner-take-all. The actual structure is a horizontal layer capturing the early stage of the search, with vertical layers underneath that defend different parts of the eventual decision.
The supply-owning bet is being made, by players who already had the assets and now have the urgency.
The corporate bet is in motion. The TMCs and newer entrants are building, at different speeds and with different conviction.
The curation, destination and criteria bets are mostly open. The foundations exist. No one has moved decisively.
For everyone outside the five — most independent hotels, most mid-tier chains, most operators without a network or a destination authority behind them — the bet is not available at all. The question is which vertical to be findable inside.
by Markus Busch, Editor/Publisher Hospitality.today
Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →