Will Agentic Hotel Bookings Flourish or Fizzle?

Ask Alibaba, Not Google.

200 million. That's how many AI-native orders ran through Alibaba's Qwen app during Spring Festival alone. Fliggy's AI-driven bookings jumped 800% year-over-year in that same window, with attraction-ticket conversion now matching OTAs that have run for two decades. Real revenue, banked.

Meanwhile, half the panels I've sat on this year are still relitigating "Book on Google," debating whether agentic booking could even work. Cute. Alibaba skipped it and went straight to collecting the receipts.

Google's pilot list: Booking.com, Expedia, Marriott, Hilton, IHG, Wyndham. Ten companies that have spent two decades fighting over guest data and loyalty points, each wanting the agent to book through its own system. That friction is why every booking still needs a human to click confirm.

Fliggy's flyai skill took the opposite bet: open source, free to any developer, no account required, no gatekeeping. Every developer plays the same game at once, instead of ten companies fighting over who owns the sale.

Keep calling it "agentic" if it makes the press release feel modern. In Hangzhou, agentic means the AI finishes the job. In the West, it apparently still means ten companies need to agree on whose logo goes on the receipt.

Google's agentic booking will flourish if its conversational AI collapses hours of travel research into simple, secure Google Pay transactions backed by major hotel and OTA partners. However, it will fizzle if consumer anxiety over high-cost bookings prevents users from trusting an AI agent without traditional visual validation, or if strict global regulations like the EU's DMA permanently limit its market footprint.

I believe “this time is different.” Both “Book on Google” and agentic booking are not MoR and instead route bookings to the end service providers (OTA or direct). So clearly, the question is why would travelers be comfortable “submitting” bookings directly to AI, but not BoG?

With BoG, Google tried to close the loop and become a semi-transactional layer. But users never saw its identity as a booking channel. Google Hotels is, and has always been, a discovery engine (metasearch). Google attempted to "act" like an OTA, but lacked the last-mile “trust” that OTAs have built over decades. Therefore, forcing checkout into that flow caused confusion.

Why would this time be different? My answer is: the intent differs. Travelers use metasearch to compare but they don't trust metasearch to handle their transactions. But when people use AI, they're more trusting it to execute it for them as an extension of, not in lieu of, the end service providers. Why? Because AI is infrastructure, just like UI (user interface), which lives on top of browser and app technologies. Travelers would be comfortable booking with AI the same way they already are with browsers and apps.

Agentic AI bookings will flourish, but not yet. No doubt, agentic AI will diminish the role of hotel websites, together with functionalities like booking engines and chatbots, but not yet. Not in the next at least 10 years. 

Why is that? Inertia loyalty! Consumers hate change. They prefer using the same tools, channels, platforms they are comfortable with. If for 25 years I have booked on Booking, Expedia or Marriott, why change? 

Inertia loyalty was the reason why Book on Google and Book on TripAdvisor failed miserably. Why after 30 years of commercial Internet, only 60%-65% of hotel bookings are online.

In addition, history has shown that as far as technology is concerned, there are early adopters, adopters, late adopters and non-adopters. 

In the future, I don’t see a cataclysmic explosion of AI-related bookings, but gradual blurring the lines between AI Search/Agentic AI and the traditional online distribution channels. You may use Google Search, but in reality it is Gemini AI already providing the answers. You may use your personal AI Agent to book a hotel, but the actual booking will be powered by Booking, Expedia, Marriott, Hilton, etc.

So, at the moment, hotels have to cater to two audiences: AI agents and humans using traditional channels. 

Agentic booking will flourish. First as a decision channel, then as a booking channel.

Half of travellers already say they’re ready to hand planning and booking to AI. Once the decision is delegated, booking is just the last click.

For 25 years, the funnel was visibility, price, conversion. Now it becomes experience, trust, value, selection, then booking.

For now, the launch partners are the incumbents. OTAs still control pricing and booking, making AI a new front door to the same houses.

The real shift is selection. An agent might shortlist three hotels, maybe four. Being selected becomes more valuable than being listed.

Selection will depend on trust machines can read: reviews, sentiment, product quality and operational consistency and so agentic booking will flourish but It will be least forgiving to the vanilla hotel.

I think the more important question is who the agents will book through.

A traveler might soon be able to say: “Find me a quiet hotel in Stockholm near my meetings, with a gym, late check-in and a room similar to the ones I normally prefer.” The agent can research, compare, negotiate and eventually complete the transaction.

At that point, the hotel website is no longer necessarily the interface. Neither is an OTA. The AI agent is.

That creates an enormous opportunity for hotels.

Historically, OTAs won because they aggregated supply, standardized the booking experience and made hotels easy to compare and transact with. In an agentic world, hotels can make their own inventory, rates, policies and services directly accessible to AI agents.

But just exposing availability is not enough.

Hotel systems will need to understand the guest, decide what to offer, optimize for conversion and revenue, and execute the transaction. In other words, consumer-side AI agents will probably need hotel-side AI agents to transact with.

This is why I believe agentic booking is much bigger than another distribution channel. It could potentially change hotel distribution.

After hearing a wide range of views at Destination AI, I think the debate still contains more pontification than evidence about how travelers will actually behave.

Our research with Curacity, surveying 1,000+ active U.S. travelers, found that AI serves different purposes across segments. Budget travelers seek value, business travelers prioritize convenience, and luxury travelers look for fit and reassurance. Over 60% named accuracy as their top concern. Trusting AI to help plan a trip is different from trusting it to make the purchase.

Our visibility research highlights another consequential issue: which hotels get considered. Across hotels in 13 markets, properties in the top quartile of third-party coverage were recommended approximately 3.3 times as often as those in the bottom quartile. This association does not establish causation, but it demonstrates why visibility deserves attention alongside booking functionality.

AI could reshape hotel distribution even if travelers retain control of the final transaction. The experience of Book on Google does not settle that question.

We need more experimentation and observed behavior to understand who delegates which decisions, under what circumstances, and how AI recommendations influence the hotels travelers ultimately consider and book across different travel occasions.

As with any other new technology, there will be a portion of early adopters. Growth rates will look phenomenal at the outset as they are starting from close to zero. Making agentic reservations in AI mode will certainly appeal to some of us - however, AI still requires the go-ahead by a human during many steps during the entire process. I am not sure whether this constitutes a true agentic reservation or is simply another way to "package" the booking process in the same window.

The emergence of Muse by Meta is adding a new dimension to the question. Giving Muse access credentials (via vault) to Hotel Loyalty programs opens the possibility that a user can direct Muse to book Hotel ABC in a particular city through the user's Loyalty account. The main question comes down to trust with CC / Payment information and whether a user is open to sharing that detail with agents. Most likely, the agent will require the final approval from the user before completing a transaction.

Whether users will be open to asking AI agents to research, select and book in an unknown destination requires even more trust and will take time to develop

It's a new generation of technology and we cannot afford to view it through a lens of past technical capability. There is too much overall value in the functionality and fit into a new computing experience to not see things through to a better conclusion than in the past.

Flourish - Eventually it will become natural to connect the process of AI and Agentic AI research to the final purchase process. It is a question of the timing of the full flow through of connectivity of systems to fulfil that outcome.

Today's handful of connected parties will be tomorrows all parties on a similar platform. Between now and then the entire cycle will have improved to support a smooth transaction.

My view is that this new model now requires a true shopping cart experience vs a traditional booking engine. This is both necessary but also good for the industry from a merchandising and retailing perspective.

The question that always remain for me is where the trust will lie. Will the user trust third parties' aggregators at scale to represent and recommend? Or will they want a deeper connection to the product provider after the agents have done their work?

I believe agentic hotel bookings will flourish — but their real impact will be much broader than simply moving the booking button into an AI interface.

I believe agentic hotel bookings will be huge. AI search is growing rapidly, and booking is the logical next step as travelers become more confident in the recommendations. Once an assistant understands your preferences, budget and priorities, asking it to complete the reservation becomes a natural extension of the conversation.

The comparison with “Book on Google” overlooks a fundamental difference: that initiative simplified a transaction; agentic AI can simplify the entire decision. The value lies in reducing the work of searching, comparing and choosing, then carrying that understanding through to booking.

Keep an eye on Baby Boomers as a potentially leading demographic for adoption. We often assume younger travelers will drive technological change, but a conversational assistant could be particularly appealing to people who find navigating multiple websites, room categories, rate conditions and booking forms frustrating. An assistant that explains the options clearly and handles the details offers a compelling benefit.

Trust will determine the pace: transparent pricing, accurate recommendations, control over the final purchase and accessible support when something goes wrong. As those capabilities improve, I expect adoption to accelerate considerably.

Agentic booking will likely flourish, but for travelers and Google first, not for every hotel.

Why this time differs from Book on Google: the decision now happens inside the answer, so booking there is the next step, not a detour. The major OTAs and five global chains are integrated from day one. And Google needs agentic commerce to defend search against AI assistants.

The hotel question is different: who do you reach on Google? Everyone, and therefore no one in particular. Google passes partners only the data needed to complete the booking. The hotel gets the transaction, never the reason for the trip.

That pushes the overnight stay further toward a generic product, where the lowest price wins.

Unless the hotel gives the agent something else to match. A traveler who asks for a quiet room near the congress center can be matched on fit before price. Hotels with a clear target audience must know how that audience books, and describe themselves so precisely that the agent recognizes them as the answer. Everyone else competes on rate.

Max, I think there are really two questions in one here. 1. Will Google retain (and now expand) its gatekeeper position? and 2. What will the future of Agentic AI bookings look like?

With AI development moving at a breakneck pace, no one can accurately predict what the future will look like, but it most certainly will look different from what it does today. I’ll outline one possibility below:

Imagine if we, as individual travelers, had our own (trusted) agentic AI agent. It knows our preferences, affiliations, motivations, and much, much more. Now imagine individual travel suppliers having their own AI agents. It wouldn’t be too far-fetched to envision a world where agent-to-agent negotiations happen, and a refined set of options is delivered to the traveler for final selection.

Suppliers will benefit from direct bookings from more profitable guests, and guests will benefit from knowing that their accommodation selections more completely match their needs. In this agent-to-agent scenario, discovery, booking, service delivery, and support are neatly wrapped up in a single engagement.

Disintermediation? … Most definitely!

A market segment of one? … Certainly a big step in that direction.

I believe agentic hotel bookings will flourish, but perhaps not in the way we are currently discussing them.

Comparing this with Book on Google is understandable, but there is one fundamental difference. Book on Google tried to change where the traveller completed the transaction. Agentic AI has the potential to change how the traveller makes the decision in the first place.

For hotels, the question is therefore not only whether travellers will book through AI. Will the agent find your hotel? Will it understand your product well enough to recommend it? Will it have access to the right availability, rates and content? And who will own the guest relationship and the economics of that booking?

With both major hotel groups and OTAs among the initial partners, agentic booking could create a new direct distribution opportunity, but it could just as easily strengthen intermediaries.

Hotels should prepare for a world in which customers may no longer search, compare and click in the way we have spent the last 20 years optimising for.

That is much bigger than simply adding another booking channel.

Agentic AI isn't a new booking channel. The hotel brand or OTA retains the transaction and customer-service responsibility. It's a new decision layer for the customer —understanding needs, researching, comparing and recommending hotels before directing the transaction - potentially a fundamental change to the traditional hotel booking funnel.

In effect, as customers become more AI savvy and vice-versa, Agentic AI understand requirements and can perform in seconds comparison work that previously required travelers to conduct 10–30 searches and clicks, visit multiple hotel and OTA websites, and evaluate competing properties.

That shifts the strategic battleground from simply winning the booking to first winning the comparison and recommendation.

Therefore, the critical question may not be whether travelers will book hotels through Agentic AI. The more important question is:

Will Agentic AI recommend your hotel—and, when it does, through which channel will it complete the transaction?

My belief is that routine bookings and lower-tier inventory go highly agentic. The airport hotel, the monthly one-night business trip...they'll compete aggressively on things like price, and the agent will book.

Upscale and luxury will stay a human experience. The time between inspiration, that first recommendation, and booking will still take weeks and span multiple surfaces and conversations. These are high-consideration, often emotion-driven purchases.

The key is designing a world that optimizes for each. But, both tiers have one thing in common, which is that the decision is made before the booking. In the routine tier, the agent picks from a shortlist. In luxury, the traveler builds that shortlist over weeks, increasingly from AI answers. Either way, getting AI to recommend your property is mission critical (you're either in the consideration set or not...it's binary). The next guest a hotel loses won't go to a competitor with a better rate, but to the one AI found more reasons to trust and therefore recommend.

The question is not whether travellers will ask agents to find and book hotels. It is who controls what that agent can see, recommend and transact.

In a way that’s not different than a traditional travel agent.

I don’t see ai agents as a separate booking channel. Instead they are more like a channel manager. A tool connecting supply and demand.

OTAs will dominate that layer. They already have the connectivity, customer relationships and booking infrastructure to make agentic transactions convenient.

Major hotel brands might have a chance within their own ecosystems.

Independent hotels? They face the same distribution problem they face today, only with another decisionmaker standing between them and the guest and an added layer of complexity. And the added disadvantage to not have the resources to figure this all out on their own.

Indies need to rely on third party partnerships to navigate this shift, while making sure they don’t surrender even more control over their pricing, visibility and guest relationships.

So, the technology will not fizzle. The promise that it will level the playing field for independent hotels might. The winners be those who understand distribution, make their inventory accessible and protect their commercial position before someone else defines it for them.

Let's start by asking whether the booking agent is solving a real consumer problem. I think the booking transaction is not really a problem for consumers right now. Discovery and choosing the right hotel, as well as the booking channel, are. There is also an issue of trust in AI agents. Can I trust the AI's choice? Will my financial information be handled correctly? I think AI agents will be very helpful at the planning stage, and then consumers will make the final decisions about the booking channel, hotel choice, and completing the final booking transaction.

What does it mean for hotels? Hotels need to make sure AI has correct information to discover and recommend them. There are tons of recommendations; start with your website's FAQ section.

I think AI will flourish at the discovery stage, and consumers will make decisions based on the information AI provides. I don't see “type a prompt and let AI do all the work for you” as mass adoption anytime soon. I think hotel guests will delegate some stages of the booking journey to AI, but not all. Consumers will want some control, and how much will depend on how much they trust the AI tools used. It also dependsd ot the type of traveler. A frequent business traveler rebooking a familiar brand may let the agent do everything. A family planning its annual vacation probably won't. How much people hand over will track how much they trust the tool.