AI Assistants Set to Disrupt Travel Re-Shopping

On 28 October 2026, PhocusWire reported a warning from the hotel industry about personal AI assistants. Rick Posey, vice president of product at Cendyn, said the process of re-shopping carries landmines the industry is not ready for yet. Re-shopping means booking a room, then cancelling and rebooking when a better deal appears. The assistants now entering the market can run that process on instruction, at a scale a traveller checking prices by hand cannot match.

This matters for the industry before it matters for the guest. The first number to break will be pace, the rate at which reservations are made for a particular date. Pace feeds pricing decisions, staffing plans and inventory holds. If pace looks stronger than it is, a hotel can hold a rate it should cut, or cut a rate it should hold. The distortion arrives in the measurement system, well before the booking volume that would justify a change in strategy.

How re-shopping by assistant works

Booking Holdings launched Vayo on 14 October 2026. Vayo is an AI travel startup based in Seattle, with plans to invest US$100 million, subject to product progress and market conditions. It collects travel confirmations from a user's inbox and organises them into itineraries. It then monitors bookings for price drops and availability, and alerts the traveller when rebooking at a better rate is possible.

For hotels, the current step is a prompt. Vayo can automatically capture price drops and secure refunds for some flights, PhocusWire reported. For now, if a hotel price drops, Vayo prompts the user to take additional action. That limit is temporary. Mike Fridgen, Vayo chief executive, told PhocusWire the company will roll out transactional booking for flights and hotels in the coming weeks, which will extend automated price-drop capture to more scenarios, including hotels. General availability is likely in the first half of 2027.

Vayo also works across channels. Fridgen said it can organise and monitor reservations made through competitors such as Expedia, and reservations made directly with hotels. That detail matters for a hotel that assumed a direct booking was a stable booking. A direct reservation can now be watched by the same assistant that watches an online travel agency reservation.

The other assistants work differently, and the difference shows the range of the problem. Instinct, a personal AI agent launched on an invite-only basis in August 2026, checks restaurant sites every five seconds for available slots, its founder said in an interview cited by PhocusWire. More than 50 per cent of Instinct transactions relate to travel, PhocusWire reported. Instinct raised US$1 billion in a Series C round announced on 28 September 2026, at a US$10 billion valuation, according to TechCrunch reporting. Meta's Muse is a third model, a general purpose assistant the user can instruct directly. Posey gave the example that applies to hotels: a user can instruct Muse to keep shopping for a specific rate in a specific time frame, staying under a price range.

Re-shopping is old. Automation at this scale is new

The industry has seen re-shopping before. FairFly, now rebranded as Oversee, and TripBam, acquired by Emburse, built business models on the same principle. Google, Skyscanner and Kayak have offered price tracking for many years. A revenue manager who has dealt with those tools could reasonably ask what has changed.

Three things have changed. First, the instruction is continuous. A traveller once checked a price when time allowed. An assistant checks on a schedule, or on a trigger, without the traveller opening an app. Second, the capital behind the tools has jumped. Booking Holdings plans US$100 million for Vayo. Instinct raised US$1 billion at a US$10 billion valuation within about a month of a US$2.5 billion valuation in August 2026. Third, the booking step is moving inside the assistant. Google confirmed on 7 August 2026 that it had begun a limited United States test of agentic hotel booking inside Search's AI Mode, with Amadeus as an infrastructure partner. The assistant is moving from recommending a hotel to completing a booking, and from completing a booking to watching it afterwards.

The first metric to break is pace

Posey described the mechanism in two steps. An instruction to keep shopping inflates a hotel's analytics. The look-to-book ratio rises, because the assistant looks far more often than a human traveller. Then, when bookings are created, pace is affected. Pace can look stronger than it actually is because people can instruct the AI assistant to cancel if something else pops up at a lower rate, even somewhere else.

That sequence is the argument of this piece. Look-to-book distortion is visible early. Pace distortion follows, because a booking created by an assistant carries a cancellation instruction with it. A hotel reading pace as demand will misread it. A hotel reading pace as a signal to hold rate will hold rate on bookings that were never firm.

The mechanism has a second edge. Jordi Serra, chief executive of marketing and revenue platform 123Compare.me, told PhocusWire the risk is sharpest for a revenue manager who needs to lower prices because the hotel is empty. If the bookings were on a cancelable basis, guests would cancel and take the new price. A price cut intended to fill rooms can therefore trigger cancellations of rooms already sold at the higher rate. The hotel sells the same room twice in its own data, and once in reality.

The volume is not there yet

The honest complication is volume. PhocusWire reported little evidence so far that re-shopping from these new assistants harms hotels. Hotel executives generally quote around 1 per cent for inbound referral traffic from conventional AI search platforms, as a percentage of total nights being booked. Booking Holdings chief executive Glenn Fogel said on 23 September 2026 that AI traffic is significantly below 1 per cent of room nights. Lighthouse hotel website analytics from July 2026 found AI referral traffic to independent hotel websites at roughly 1 per cent of total site visits.

The State of Distribution 2026 report, from New York University, RateGain and HEDNA, gives a different measure. It puts AI search at 4 per cent of bookings in 2026, against 34 per cent for online travel agencies and 18 per cent for hotel websites and mobile apps. Those two measures do not describe the same thing. One counts referrals and room nights through a specific platform. The other counts booking share across channels as reported by hotels. The piece uses both, with their bases stated, and does not merge them into a single AI share.

Identification is also weak. A revenue management specialist at a global hotel management company, who asked to remain anonymous, told PhocusWire it is hard to identify AI bookings, and that the company is deep diving into bookings at the moment. Safa Rahal, a commercial and AI strategist in hospitality, said capacity and volume are not quite coming in yet, although references are appearing in Google Analytics referral platforms through AI agents. At the Destination AI Summit in Washington, D.C., Posey said hoteliers were already asking how to tell an agent from a human, with speed to booking as one clue.

Cost may also limit the damage. Serra argued a service that checks prices every day is not cheap. If the agent is a personal agent, it will process tokens, and at some point the user will face a cost. That argument cuts both ways. It limits casual checking by every traveller, and it concentrates checking among travellers who value the saving enough to pay for it, which is the group most likely to cancel and rebook.

Why that may not matter much

1. Pace distortion starts before booking volume does. An assistant can inflate look-to-book with repeated checks while creating few bookings. A hotel that waits for AI bookings to reach a material share before fixing its measurement will already have read distorted pace for months.

2. The questions have already moved. Posey said that 12 months ago, referrals from ChatGPT were a question mark for many hoteliers. Now more people are tuned into it, and hoteliers are asking good questions about preparation. A measurement problem that is discussed at an industry summit is a problem a revenue team can start tagging now, by channel and by cancellation, before the volume arrives.

3. Pace decisions are long-lived. A rate held on false pace, or a rate cut that triggers cancellations, affects revenue for a stay date that cannot be re-sold after the date passes. The cost of a wrong pace reading is therefore higher than the current AI booking share suggests. A small channel can cause a large pricing error if the pricing decision uses the channel's bookings as a demand signal.

The real test is named here. When Vayo rolls out transactional hotel booking, and when Instinct moves beyond invite-only access, hotels should be able to report cancellations by assistant channel. If agent-tagged cancellations stay near zero, the pace warning was early. If they rise with transactional volume, the warning was on time. Nobody can answer that yet, because the transactional step for hotels is still rolling out.

What to watch

Hotels should watch three things, each checkable against a named source. First, Vayo's hotel transactional rollout, promised in the coming weeks from October 2026, with general availability likely in the first half of 2027. That is the moment automated price-drop capture extends to hotels in the way it already applies to some flights. Second, the next State of Distribution report. Watch whether AI search moves from 4 per cent of bookings, and whether the share of hotels reporting no change or only a minor change to distribution strategy falls from 55 per cent. Third, agent identification itself. Rahal predicted hotels will start to look at AI agents the way they look at online travel agencies, with credibility based on minimal cancellations. A hotel that can tag an agent booking, and report its cancellation rate separately, will know whether its pace is real.

The industry has learned this sequence before. Every new channel, from online travel agencies to social media to review sites, was first treated as something to block, Rahal told PhocusWire. The lesson she draws is to work with the channel. For pace, working with it starts with a narrower task: separate the bookings an assistant is still shopping from the bookings a guest has actually settled on, before pace is used to set a price.