The erosion of Booking.com

The OTA built a $170 billion business by controlling what travelers see. That control is fragmenting across a dozen AI platforms — and management is pretending otherwise

Key takeaways

• Booking.com's real business is selling visibility, not processing bookings. Commissions are just the starting point. Preferred Partner fees, Genius participation, and sponsored placements extract the real margin. Remove interface control, and the margin structure collapses.
• AI platforms are taking control of discovery — and will monetise it themselves. Google, OpenAI, Perplexity, Microsoft, and Amazon are all building agentic travel booking. AI recommendations will still have visibility hierarchies worth monetising. But those platforms will collect the payment, not Booking.com.
• The "partnerships" are capitulation. Booking.com provides inventory; the platforms control discovery and recommendations. Hotel brands connect directly through their CRSs. Independent hotels connect through channel managers and GDS providers. Neither needs Booking.com as the aggregation layer for AI distribution.
• The market hasn't priced this risk. At 34x earnings, the valuation assumes continued growth and sustained margins — not visibility economics migrating to AI platforms within five years.

The merchandising machine

To understand what Booking.com stands to lose, you have to understand how it actually makes money.

The company doesn't simply charge hotels a commission for bookings. It charges them for placement. The entire interface is a monetisation engine designed to extract maximum revenue from the gap between what travelers see first and what they see last.

Base commission rates run from 15% to 25%, with most properties paying around 15%. But that's just the starting point. Hotels that want better visibility pay more. Preferred Partner status costs additional commission points. Participation in the Genius loyalty program requires funding discounts. Sponsored placements cost extra. Mobile-first positioning costs extra. Every pixel of the Booking.com interface has a price.

This is why Booking Holdings generates operating margins that dwarf traditional travel businesses. The company isn't in the hotel business. It isn't even really in the booking business. It's in the visibility business — auctioning placement to suppliers desperate for exposure.

In 2024, Booking Holdings spent $7.3 billion on marketing, roughly 31% of revenue. That spend only makes sense because the company controls what happens after travelers arrive. The investment acquires attention; the interface monetises it. Remove that control, and the entire economic logic collapses.

The interface is the product

When a traveler searches on Booking.com, they encounter a carefully engineered sequence of choices. Default sort orders favour higher-commission properties. "Recommended" labels go to partners who pay for them. Price displays are structured to make comparison difficult. Review summaries emphasise what Booking wants emphasised.

None of this is hidden — it's the explicit value proposition Booking sells to hotels. Pay more, appear higher. Fund discounts, get the Genius badge. Accept restrictive rate parity terms, get preferential treatment. The entire supplier relationship is built on the premise that Booking.com controls traveler attention and will sell access to it.

This is not a service business. It's an advertising business that happens to process transactions.

The margins reflect this. Booking engine, customer service, and payment processing are valuable but commoditised functions. Controlling what travelers see — and charging suppliers for better placement — is where the exceptional margin lives. Booking.com has built both, but investors have valued it based on the latter.

The payments layer

There's another margin layer most travelers never see: payments.

Booking.com has aggressively shifted from an "agency model" (where travelers pay hotels directly) to a "merchant model" (where Booking.com collects payment upfront and pays hotels later). Merchant revenue now accounts for roughly 60% of total revenue — up from under 35% just a few years ago.

This shift unlocks multiple revenue streams beyond commissions.

First, there's the float. When a traveler books in January for an April stay, Booking.com holds that money for months before paying the hotel. Multiply this across a billion room nights annually, and the interest income is substantial.

Second, there are payment processing fees. Booking.com pays hotels through Virtual Credit Cards, which cost properties around 3% to process. Booking.com almost certainly receives rebates from card networks on this volume — an entirely incremental revenue stream that doesn't appear in commission disclosures.

Third, there's foreign exchange. Travelers prefer paying in their home currency. Hotels operate in local currency. Booking.com sits in the middle, capturing the spread.

None of this requires Booking.com to control discovery. But all of it requires Booking.com to be in the transaction flow. If AI platforms route bookings directly to hotels — or if hotels insist on direct payment relationships to avoid the fees — this margin layer disappears along with the visibility economics.

The fragmentation of discovery

A year ago, the threat to Booking.com looked like it might consolidate around one or two dominant AI assistants — ChatGPT or Google's Gemini. Whichever platform won the AI race would control travel discovery.

The reality is messier — and worse.

In January 2025, OpenAI launched Operator, an AI agent that can browse the web, make reservations, and complete bookings autonomously. Booking.com, Tripadvisor, Priceline, and Hipcamp partnered at launch. By July, OpenAI had rolled out ChatGPT agent mode to paying subscribers, enabling the chatbot to navigate websites, compare options, and book hotels without human intervention. In October, OpenAI unveiled its Apps SDK, turning ChatGPT into what CEO Sam Altman called "an AI operating system" — with Booking.com and Expedia as launch partners, accessing ChatGPT's 800 million weekly active users.

Meanwhile, Perplexity partnered with Selfbook and Tripadvisor to offer hotel booking directly within its AI search engine. The company's CEO declared 2025 the "year of agentic action" — the idea that users should be able to book anything from the point of inspiration. Tripadvisor's executives reported that customers arriving via Perplexity are "high intent" and convert at rates above typical traffic.

Microsoft integrated Expedia with Copilot Actions for enterprise users. Amazon launched Alexa+ with travel booking capabilities. Google continued building agentic booking for Gemini, working with Booking.com, Expedia, Marriott, IHG, and others.

The threat isn't a single dominant platform. It's a dozen platforms, each demanding inventory access, each controlling its own interface, each deciding independently what "recommended" means.

What happens when discovery moves upstream

Agentic AI changes the location of the decision — and, crucially, who controls it.

When a traveler asks any of these AI assistants to find and book a hotel, the selection happens inside the model. There may still be a ranked list of options. There may still be a "recommended" choice. There may still be prominence given to certain properties over others. Visibility hierarchies don't necessarily disappear.

But Booking.com won't control them.

Each AI platform will develop its own monetisation model. Perhaps suppliers will pay for preferential placement in recommendations. Perhaps "sponsored" options will appear first in conversational responses. Perhaps the default suggestion — "I found the Marriott Downtown, should I book it?" — will go to the highest bidder.

Whatever models emerge, they will belong to Google, OpenAI, Perplexity, Microsoft, and Amazon — not Booking.com. The company that spent two decades perfecting the art of selling interface placement will find itself on the other side of that transaction — paying for visibility rather than selling it.

This is the structural shift that matters. The question isn't whether visibility can still be bought in an agentic world. The question is who collects the payment.

Partnerships that aren't

Booking.com's response to this fragmentation has been to partner with everyone. The company is working with Google on Gemini. It partnered with OpenAI on Operator and the ChatGPT Apps SDK. It's feeding inventory to whatever AI platform will take it.

Management presents this as strategic positioning. In reality, it's capitulation.

Look at what these partnerships actually involve. Booking.com provides inventory data, pricing, and booking infrastructure. The AI platforms provide the user relationship, the discovery interface, and the recommendation logic. Booking.com becomes a backend supplier; the platforms become the frontend.

Hotel brands connect directly to these platforms through their CRSs. Independent hotels connect through channel managers and GDS providers. Neither needs Booking.com as the aggregation layer for AI distribution.

Booking.com's invitation to participate in these platforms isn't a sign of strength. It's a sign that AI companies want optionality — multiple inventory sources that can be played against each other. The more suppliers competing to feed each platform, the more leverage the platform has over all of them.

Every partnership announcement is a press release celebrating Booking.com's diminished role.

Who controls the visibility hierarchy

Here's the core problem Booking.com faces: even if visibility gradients persist in agentic AI, the company won't own them.

Consider how Booking.com makes money today. A hotel pays 15% base commission plus additional fees for Preferred Partner status, Genius participation, and sponsored placement. Booking.com pockets the premium because it controls the interface. The company decides what "recommended" means, what appears first, what gets the trust badges.

Now consider how ChatGPT or Gemini might work. A traveler asks for hotel recommendations. The model returns three options, with one marked as "Top Pick." That designation has value — suppliers would pay for it. But OpenAI or Google, not Booking.com, decides what earns the label. They set the criteria. They collect the premium.

Booking.com could theoretically pay these platforms for preferential treatment of its inventory. But that inverts the entire business model. Instead of collecting visibility fees from suppliers, Booking.com would be paying visibility fees to AI platforms — while still paying suppliers their contracted rates. The margin structure doesn't just compress; it potentially inverts.

OpenAI already monetises ChatGPT through subscriptions and is building commerce capabilities. Google already monetises AI Overviews in search. Perplexity is experimenting with transaction fees. There's no reason to think any of these platforms will leave visibility economics on the table.

The question is whether Booking.com will be selling access or buying it — across not one platform, but a dozen.

The erosion of marketing returns

Booking Holdings has spent decades building competitive moats through marketing investment. The cumulative spend is staggering — $7.3 billion in 2024 alone, $6.8 billion the year before, similar figures stretching back years. The company dominates Google search results, runs perpetual brand campaigns, and has achieved near-universal name recognition among travelers.

This investment is optimised for a specific user behavior: travelers who search Google for hotels, or who navigate directly to Booking.com's app or website. SEO dominance pays off when travelers type queries into search engines. Brand awareness pays off when travelers choose Booking.com over Expedia. App install campaigns pay off when travelers open the Booking.com app to start their search.

The threat from AI isn't that this marketing becomes misdirected. It's that the behavior it targets becomes less common.

If travelers increasingly say "find me a hotel in Barcelona" to their default AI assistant rather than searching Google or opening a booking app, the entire journey bypasses Booking.com's interface. The marketing spend doesn't acquire those users — not because it's poorly targeted, but because those users never enter the funnel it's designed to capture.

This doesn't mean Booking.com's brand becomes worthless. Travelers may still prefer completing transactions through a name they recognise. But brand preference is a weaker asset than interface control. If Gemini or ChatGPT recommends a hotel and offers to complete the booking through Booking.com or directly with the property, brand becomes one factor among many — not the decisive chokepoint it is today.

The $7.3 billion annual marketing budget is built for a world where travelers actively seek out booking platforms. That assumption is what's at risk.

The Model Context Protocol problem

There's a technical development that should worry Booking.com more than any single partnership: the Model Context Protocol.

MCP is an open standard, introduced by Anthropic in late 2024 and quickly adopted by OpenAI, Microsoft, and Google. The idea is simple: instead of building separate integrations for each AI platform, a company builds one MCP connector that works with all of them. One integration, many AIs.

For hotels, this could be transformative. A property management system or channel manager could build a single MCP connector that makes its inventory available to ChatGPT, Gemini, Claude, and Copilot simultaneously. Hotels wouldn't need to go through Booking.com to reach AI assistants — they could connect directly.

If hotel technology vendors build these connectors, independent properties could retain direct control in the AI era. If OTAs dominate this layer instead, hotels face another round of intermediation — but the intermediary might not be Booking.com.

The protocol is still early. But the trajectory is clear: the infrastructure is being built for a world where AI platforms can access hotel inventory without going through traditional OTAs. Booking.com's role as an aggregator becomes optional rather than necessary.

The narrative gap

Listen to Booking Holdings' earnings calls and you hear a company that views AI as an opportunity.

CEO Glenn Fogel talks about using agentic models to improve how customers discover and use their platforms. The company is investing in AI-powered trip planning. Management frames the shift as a channel evolution — something to master, not something to fear.

This framing has a critical flaw: it assumes Booking.com will either control its own AI interface or maintain pricing power as a supplier to others.

If travelers use Booking.com's app to interact with an AI assistant, the company retains its position. It can build visibility monetisation into its own agent. It can maintain the supplier relationships that generate margin.

But if travelers use ChatGPT, or Gemini, or Perplexity, or Alexa, or Copilot, or whatever assistant becomes default on their devices — the interaction happens on someone else's terms. Booking.com becomes a backend supplier, not a frontend controller.

And the evidence suggests travelers will use whatever assistant is most convenient. ChatGPT has 800 million weekly active users. Gemini is embedded in Android, Chrome, and soon Siri. Alexa is in millions of homes. Booking.com's app, however popular, can't match that distribution.

Management is betting that their own AI investments will keep travelers in their ecosystem. They're not explaining what happens if travelers simply ask their default assistant instead.

What the platforms actually want

Google, OpenAI, Microsoft, and Amazon have never been protectors of suppliers. They monetise access to decision-making moments. Their business models depend on inserting themselves between intent and action, then capturing value from that position.

In traditional search, Google extracted value by charging travel companies for clicks. Booking.com and Expedia paid billions annually for Google Ads placement. The companies grumbled about the cost but accepted it because Google controlled the attention they needed.

Agentic AI raises the stakes. Instead of charging for clicks that lead to someone else's interface, these platforms can own the entire transaction flow. The user never leaves the conversation. The booking completes inside the agent.

In this model, AI platforms don't need Booking.com to succeed. They need Booking.com to compete — against Expedia, against hotel chains, against anyone else willing to provide inventory. The more suppliers feeding each platform, the more the platform can optimise for its own interests: better prices for users, better margins for the platform, commodity returns for everyone in between.

Booking.com is being invited to participate in systems designed to commoditise its role. The partnership announcements aren't votes of confidence. They're onboarding notices.

The independent hotel problem

Booking.com's defenders argue that the company provides unique value by aggregating long-tail inventory. Independent hotels need distribution, and Booking.com offers global reach they can't achieve alone.

This is true today. It may not be true tomorrow.

The same structural forces that threaten Booking.com also create opportunity for alternatives. GDS providers, channel managers, and property management systems can build MCP connectors that feed inventory directly to AI agents. A hotel with clean data and competitive pricing can surface in ChatGPT or Gemini regardless of its relationship with Booking.com.

The aggregation value Booking provides is real but not irreplaceable. It depends on Booking.com being the most efficient path to traveler attention. Once attention routes through AI agents, efficiency gets redefined — and the definition may no longer favour companies that extract 10-25% for placement services.

Independent hotels currently tolerate high commission rates because the alternative is invisibility. AI agents offer a different alternative: visibility based on data quality and price competitiveness, mediated by platforms other than Booking.com.

Whether that's better for hotels is unclear. But it's almost certainly worse for Booking.com.

What the market isn't pricing

Booking Holdings trades at roughly 34 times earnings. The valuation implies continued growth, sustained margins, and durable competitive position. It does not appear to price any meaningful probability that visibility economics migrate to AI platforms within five years.

This seems like a mistake.

The bull case requires believing that Booking.com will either control its own AI interface or maintain pricing power as a backend supplier across a fragmented landscape of competing platforms. Neither assumption is safe.

Building a competitive AI assistant requires more than investment — it requires distribution. Google has Android, Chrome, and potentially Siri. OpenAI has 800 million ChatGPT users. Microsoft has Copilot embedded in Windows and Office. Amazon has Alexa. Booking.com has an app that travelers open when they've already decided to book. The starting positions aren't comparable.

Maintaining pricing power as a multi-platform supplier requires leverage that Booking.com may not have. If Expedia, Marriott, and a dozen hotel chains all feed inventory to every major AI platform, what makes Booking.com's supply special? Scale matters less when the aggregation happens at the platform layer rather than yours.

Management's response to this threat has been to announce AI investments and highlight partnerships across every platform. Neither addresses the core problem: in an agentic world, whoever controls the conversation controls the economics. That won't be Booking.com.

The bottom line

Booking.com built one of the most successful businesses in internet history by recognising a fundamental truth: whoever controls what travelers see controls the economics of travel. For two decades, the company executed that insight brilliantly — investing billions to acquire attention, then monetising that attention through carefully optimised interfaces.

Agentic AI doesn't eliminate visibility hierarchies. It transfers them.

When discovery, comparison, and recommendation happen inside AI models — whether Google's, OpenAI's, Perplexity's, Microsoft's, or Amazon's — those platforms decide what prominence means and who pays for it. Booking.com can participate as a supplier, but it cannot control the terms. The company that perfected the art of selling placement becomes a customer buying it.

What remains for Booking.com is transaction execution and brand affinity — real assets, but not the assets that justify current margins. Processing bookings and providing customer support are utility functions. They don't command 15-25% take rates.

The AI platforms aren't inviting Booking.com to be a strategic partner. They're inviting Booking.com to compete for the right to fulfill demand they control. And they're extending that same invitation to Booking.com's competitors and suppliers simultaneously.

The company's management is treating a transfer of power as a channel opportunity. The market hasn't yet priced the possibility that visibility economics fragment across a dozen AI platforms within five years — none of which Booking.com controls.

It should.

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 →