This battered travel stock is seen as an AI victim. But the data suggests otherwise
HS 09/09/2026
Booking Holdings trades at 15.3 times forward earnings, significantly below its 10-year average of 22.3 times, even as the company keeps delivering on its long-term growth targets. The AI fear that has weighed on its stock is measurably absent from its results: traffic reaching Booking from large language models remains below 1% of room nights, and management stated, citing third-party data, that the vast majority of LLM travel referrals that do occur route to Booking. In addition, direct bookings via Booking’s own channels keep growing. A two-decade flywheel of supply, payments and loyalty, anchored by millions of properties an AI agent cannot reach on its own, grows harder to replicate every quarter. As investors divvy up which companies win or lose in the age of artificial intelligence, it’s begun to look like the crowd sees online travel company Booking Holdings as a casualty. Since Google unveiled agentic travel booking late last year, the parent of Booking.com, Priceline, Agoda, Kayak and OpenTable has traded like a company on the wrong side of a platform shift. Shares have lagged the S & P 500 and the stock is priced at a discount to other large-cap internet marketplaces. Booking’s own results tell a different story. The company just beat the high end of its guidance on every key metric, and did so in the midst of a Middle East conflict that has weighed on the travel sector most of its fiscal year. Meanwhile, the threat from AI has only shown up in Booking’s thinnest business, and arguably the company has structural advantages that will protect its core business from encroachment. Why the stock weakness? Three overhangs have weighed on the stock. The first is AI. When Google expanded AI Mode’s travel capabilities last November, the online travel agencies sold off on fears that chatbots would capture the attention of vacationers and squeeze the middlemen. Booking gave the bears a real data point in the third quarter: a $457 million goodwill impairment at Kayak, its metasearch brand, explicitly citing rising customer acquisition costs as AI answers absorb search traffic. The second is geopolitics. The Middle East conflict hit travel bookings almost immediately after it began, through cancellations, rerouted flights and higher airfares, and management has been conservative in assuming it will persist. The third overhang is widely missed. Booking reports in dollars but earns mostly in other currencies, and a weak dollar flattered early-year headline growth by several points. Reported growth then decelerated sharply, but in constant currency the slowdown was modest. Much of the “collapse” that spooked the market is an exchange-rate illusion. The AI threat: What the data shows The fear embedded in the sell-off is disintermediation: Travelers ask a chatbot to book a trip, and Booking is cut out of the transaction. Traffic reaching Booking from large language models, paid and unpaid combined, remains below 1% of room nights, and management says it has barely moved. Meanwhile, the channel AI was supposed to erode keeps strengthening. The share of bookings that come to Booking directly has held firm and kept growing in absolute terms. By Morgan Stanley’s estimates, the mix was rising before ChatGPT existed and has continued rising through the AI era, led by the mobile app, which now accounts for the majority of room nights. For hotel bookings, the company makes money in two ways. Under the agency model, the traveler pays the hotel directly and Booking collects a commission after the stay. Under the merchant model — now nearly three-quarters of bookings — Booking is the merchant of record: It takes the traveler’s payment itself, keeps its commission plus fees for payment processing and currency conversion, and passes the rest to the hotel after check-in. There is a structural reason for the gap: Planning a trip and completing one are different businesses. An AI agent can assemble an itinerary, but executing bookings at global scale requires supporting payments across more than a hundred methods, compliance across more than 220 countries and territories, live inventory and someone to call when a flight is canceled at 2 a.m. Google and OpenAI have both shown they do not want to be the merchant of record — the party carrying payment risk and customer-service liability — so their early agentic products hand the checkout back to the online travel agency, and Booking is a launch partner in both. That converts the supposed threat into a referral channel. The more bookings AI initiates, the more transactions Booking settles. The company is also playing offense. Priceline’s AI assistant, Penny, can now take a traveler from conversation to completed booking. Booking.com is testing an AI discovery experience that turns a destination idea into a bookable itinerary. Management’s stated goal is that travelers should get everything inside Booking’s own apps that a chatbot offers, with the ability to book the reservation. The bet is the best defense against AI capturing its customers is owning the AI they use. The honest read on Kayak is that AI genuinely is eating thin, top-of-funnel discovery businesses, and that is precisely what Kayak is: metasearch traffic that clicks through to book elsewhere. Booking.com, however, operates the fulfillment layer beneath it. So far, AI has cut Booking’s costs rather than its business. Customer-service cost per booking is falling at a double-digit rate, and personnel expenses were nearly flat last quarter even as revenue grew. A flywheel two decades in the making Booking’s real product is not a website. It is a self-reinforcing loop that gets harder to copy every quarter it spins. The core wheel is the classic two-sided marketplace at a scale no travel rival matches: More than 4.7 million properties attract travelers, and more than a billion room nights a year of demand attract properties. What makes Booking’s version unusually defensible is whose supply it is. The overwhelming majority of its room nights come from alternative accommodations, independent hotels and small chains, not big brands, and much of that inventory connects to Booking directly rather than through third-party software. That long tail is precisely the supply an AI agent cannot reach on its own. A chatbot can find a Marriott ; it cannot contract, price and guarantee a family-run guesthouse in Puglia, Italy. The payments layer welds the wheel shut. With merchant bookings, the company sits inside the transaction, handling processing, currency conversion, refunds and fraud for small hotels that could never build that infrastructure themselves. CEO Glenn Fogel calls payments “the glue” of its business, and that arguably undersells it: It is the piece a would-be rival would need years to replicate. Then the loop closes at the customer. The top tiers of its Genius loyalty program represent about a third of active customers but the majority of room nights. Travelers who add a second vertical, such as a flight, a car or an attraction, return more often, book more direct and cancel less. Each traveler acquired through paid channels becomes a direct, repeat, app-based customer who costs nothing to reacquire. The hotels feel the pull too: Operators cannot leave the platform without losing business. The valuation gap Management publishes what it calls a long-term growth algorithm, its standing target of at least 8% annual growth in gross bookings and revenue in constant currency and 15% growth in earnings per share. In the worst macro quarter of the year, results still tracked to those growth targets: gross bookings growth on target, margins expanded and adjusted EPS up 15%. At the high end, full-year guidance still matches the original outlook in constant currency, war included. Since the quarterly report, shares have been volatile and were trading at 15.3 times forward earnings against a 10-year average of 22.3 times — a gap that implies roughly 46% upside from reverting to their own history alone, before any earnings growth. This is what a valuation dislocation looks like: Booking has underperformed the S & P 500 by roughly 37 percentage points over the past year even though its latest reported results show little evidence of a comparable deterioration in the business. In Q2, room nights grew 5%, gross bookings rose 9%, revenue increased 8% and adjusted EBITDA grew 9% to $2.6 billion, producing a roughly 36% margin. The stock is not without risk. First, there remains a threat of re-escalation in the Middle East, which would hit bookings again. The company only records revenue when travelers check in. The mitigant is that the company has already absorbed most of a year of this conflict and still guided to its original full-year outlook. Travel demand also has historically snapped back quickly once disruptions end. The second risk is Google and AI eroding the free search traffic Booking has long relied on, which is what forced the Kayak write-down. The mitigant is that losing the top of the funnel is not the same as losing the booking. The early AI travel agents still hand the checkout to Booking, which remains the merchant of record, so even trips planned in a chatbot are completed and paid for on its platform. On top of that, most bookings now come through the company’s app and website —and that share is still growing. AI-driven bookings remain below 1% of room nights. Bottom line Booking Holdings is being priced as a legacy intermediary awaiting disruption. The evidence, from rising direct mix and stable commissions to a payments and supply moat that agentic products are partnering with rather than bypassing and an earnings algorithm that just executed through a war, describes the opposite: the infrastructure layer of global travel temporarily on sale for standing next to a narrative. The disruption story has the attention. The transaction economics belong to Booking, and the price does not yet reflect it. THIS CONTENT IS PROVIDED FOR INFORMATIONAL PURPOSES ONLY AND DOES NOT CONSTITUTE FINANCIAL, INVESTMENT, TAX OR LEGAL ADVICE OR A RECOMMENDATION TO BUY ANY SECURITY OR OTHER FINANCIAL ASSET. THE CONTENT IS GENERAL IN NATURE AND DOES NOT REFLECT ANY INDIVIDUAL’S UNIQUE PERSONAL CIRCUMSTANCES. THE ABOVE CONTENT MIGHT NOT BE SUITABLE FOR YOUR PARTICULAR CIRCUMSTANCES. BEFORE MAKING ANY FINANCIAL DECISIONS, YOU SHOULD STRONGLY CONSIDER SEEKING ADVICE FROM YOUR OWN FINANCIAL OR INVESTMENT ADVISOR. Click here for the full disclaimer.
