The new front door
to travel.

From filters to intent

Generative AI is beginning to change the starting point of search. Instead of requiring a complete specification, natural language travel search can work with a partial idea, ask a useful question and turn the answer into constraints.

This is already becoming a consumer behavior. Phocuswright’s 2026 research documents growing use of AI across trip planning, booking and assistance during travel.¹ Adoption does not mean every traveler wants a conversation for every task. Someone repeating a familiar trip may still prefer a compact form.

The more consequential change is who does the translation. Software can take on more of the effort of turning a human request into a search. Good interfaces should make the resulting interpretation visible and easy to correct.

Voice isn’t the revolution. Language is.

“Find me a flight from Toronto to Miami next Thursday after 3 PM, preferably Air Canada, with one checked bag, and keep it under $700.”

A conventional interface asks the traveler to distribute that sentence across several controls. A language interface reverses the interaction: listen first, understand, structure, then search.

Voice can make this particularly natural on mobile, where moving between date pickers, keyboards and filters takes attention. Typed language offers the same expressive power when speaking is inconvenient, inaccessible or simply unwanted. Voice travel search is one way into the experience, not its entire purpose.

The real advance is understanding relationships: a preferred airline is different from a required one; a departure after 3 PM is different from an arrival after 3 PM. Speech recognition alone does not solve either problem.

AI travel search still has to be travel search

Flexible language interpretation must meet exact commercial rules. “Next Thursday” needs a calendar date. “Toronto” needs an appropriate airport interpretation. A dollar amount needs a currency. A system should clarify ambiguity before presenting an assumption as fact.

“Under $700” is a ceiling; “around $700” leaves room for a tradeoff. Checked baggage is not carry-on baggage. A round trip must retain a valid offer with its outbound and return journeys, rather than combining attractive segments into an imaginary fare.

Live inventory determines prices and availability. Fare conditions, baggage inclusions and cancellation terms must come from the relevant travel systems. An articulate answer is not evidence that a flight or room exists at the quoted price.

“AI should interpret the traveler. Trusted travel systems should determine what can actually be sold.”

That boundary makes AI flight search and AI hotel search useful in commerce. When nothing matches, the right response is an honest explanation and a proposed compromise, with the traveler deciding which constraint to relax.

Conversation changes search again

The first request is only the beginning. Seeing options helps people discover their priorities. An early departure may stop looking attractive once the transfer time becomes clear.

“Only show Air Canada.” “Actually make it Friday.” “Keep it below $800.” “I need two checked bags.” Each request should update the relevant preference while preserving everything the traveler has not changed.

Context is the hard part. A new date may require a fresh inventory search. A second checked bag may change the total cost. The experience should explain those consequences without asking someone to rebuild the trip. Maintaining that thread may matter more than whether the opening sentence was spoken or typed.

Beyond inspiration

Much of the first wave of generative AI travel experimentation focused on destination ideas and itineraries. The next challenge reaches into live inventory, comparison, transactions and servicing. Recent announcements show different parts of that journey becoming connected.

On August 27, 2026, Google announced hotel booking in AI Mode, beginning a U.S. English rollout with integrated partners. The hotel or booking platform remains the merchant of record and handles customer service. Its flight experience still sends travelers to an airline or booking platform to purchase.⁵

KAYAK’s May 2026 Ask AI announcement describes a conversation alongside flight, hotel and rental-car results that update live. Travelers can also switch back to conventional controls.⁶ Expedia Group’s July 31 acquisition of Layla brings conversational planning capabilities together with its travel supply and booking technology; the announcement sets out a strategy spanning inspiration through support.⁷

Booking.com’s October 2025 announcements extended AI into property messaging and traveler support, including voice assistance and handoffs to people when needed.⁸ These developments are at different stages and have different availability. Together, they suggest that conversational travel is becoming part of the operating experience, beyond generating suggestions.

Trust may determine who wins

Comfort with asking AI for ideas does not automatically become permission to spend. Travel involves real money, border crossings and commitments that can be difficult to unwind. Cancellations, disruptions and refunds test a service after the initial recommendation has been forgotten.

Expedia Group’s 2026 research reports that 68% of travelers surveyed prefer booking with a trusted travel brand over an AI platform.⁴ That preference can coexist with interest in AI assistance. A traveler can want a better interface and still want a known organization accountable for the booking.

The distribution side

There is an infrastructure question beneath the interface. IATA’s October 2025 look-to-book paper addresses rising search volumes and the pressure they put on airline systems and costs. It also considers how generative and agentic AI affect offer creation and search demand.⁹

Look-to-book describes the relationship between shopping requests and resulting bookings. Easier conversation could encourage more variations: another departure airport, three different weekends, a longer stay, an extra bag. One casual sentence can imply many combinations.

The response cannot simply be to issue every conceivable query. Systems need to distinguish exploration from a specific purchase request, identify useful constraints and decide when clarification would be more productive than another search.

Understanding intent well enough to issue better searches may become as important as simply issuing more searches.

That is a design and commercial judgment as much as a technical one. Success should include the relevance of the choices returned and the resources required to produce them. More activity alone does not mean a better customer outcome.

Where Zippi fits

At Zippi, we’re exploring what travel looks like when natural language becomes the interface. Describe a full trip naturally, search packages, flights, and hotels, then refine the results conversationally. The goal is to organize the trip around the traveler’s intent.

“Plan me a week in Bogotá starting October 15th from Toronto and keep the whole trip under $2,500.”

A package-search request like this expresses a shared objective for flights and accommodation, with the possibility of other components over time. Turning it into a useful search means resolving the dates, traveler count, currency and scope of the budget, then comparing workable combinations against live availability.

It does not mean promising an available package at that price, or treating separate bookings as one protected purchase. The experience must be clear about what is included and how each component is booked and supported. Zippi’s package, flight, and hotel search remains in Private Beta.

We don’t think the future of travel is a chatbot pasted onto a booking engine.

We think the interface itself is changing.

From search engines to travel agents, again

Before online travel, someone could tell an agent what they wanted. The agent understood, searched, presented choices, received feedback, refined and booked. That interaction made room for uncertainty and judgment.

AI could make more of that interaction available instantly and at scale, connected to live travel infrastructure. Human expertise and established booking tools will still have a place. The opportunity is to reduce the work between expressing a need and finding a trip that meets it.