Wyndham Hotels & Resorts is placing its own agents directly inside ChatGPT, Claude, and Gemini, feeding the models real-time rates, availability, room types, and loyalty benefits - then using AI to monitor how travelers phrase their searches and adjusting its data accordingly. CEO Geoff Ballotti detailed the strategy in a Skift interview ahead of the Skift Global Forum, marking the most concrete generative engine optimization (GEO) tactic documented in the hotel sector to date. The company reports that 35% of calls to its Wyndham AI Concierge convert to bookings, and its Wyndham Connect guest engagement platform, now in more than 5,000 hotels, generated $11 million in upsells in 2025.
From hoping to be recommended to placing agents inside the models
Ballotti described an approach that moves beyond waiting for AI models to cite the chain's properties. Instead, Wyndham supplies structured data - guaranteed rates, room categories, and loyalty perks - directly to the major AI platforms. The chain then monitors long-form traveler queries and adjusts the data it feeds the models in response. "We use AI to watch AI," Ballotti told Skift. The tactic shifts the chain from passive visibility to active presence inside the assistants where travelers increasingly begin their searches.
The results are measurable, though the figures come from the company itself in a promotional interview ahead of its industry's flagship conference. Wyndham reports franchisee retention near 96%. Ballotti also flagged a problem AI does not solve: frontline staff turnover in the sector runs at 97%.
Google's Hotel Ads deadline hits independent hotels in 13 days
Google confirmed it will end third-party rates on Hotel Ads on September 30, 2026. Any Hotel Ads campaign - or Performance Max for Travel Goals campaign relying on Hotel Ads - that depends on a rate automatically sourced by Google will stop serving across the entire Hotel Ads inventory after that date. That includes both the Booking Module and Travel Promotion Ads. The only workaround is connecting a direct price feed through Hotel Center or a certified connectivity partner. Performance Max campaigns continue to serve on Search, Display, and YouTube - only the Hotel Ads portion is affected.
For independent hoteliers running Hotel Ads without their own price feed, this is the most immediately actionable item in the current landscape. Check within the next 13 days whether your campaigns rely on a Google-sourced rate, and if so, activate a direct feed before the cutoff.
The gap between AI adoption and real impact widens
The third edition of the "State of Distribution" report, produced by NYU's Tisch Center with RateGain and HEDNA, surveyed more than 270 brands and 58,000 properties across 53 countries. More than half of hotels now use generative AI or are acquiring it. Fewer than one in ten reports a reduction in manual work exceeding 30%. Most see modest gains rather than organizational change.
The trust ceiling is the bottleneck. Hotels accept AI as an assistant but remain reluctant to let it make decisions autonomously. Meanwhile, AI-generated search already produces measurable bookings, yet 55% of hotels have made no change - or only minor changes - to their distribution strategy in response. OTAs continue capturing nearly double the bookings of hotel direct channels, even though direct booking infrastructure is now nearly universal.
Separate data from Adyen and Booking.com shows the same tension among travelers. French consumers' use of AI for destination discovery jumped 74%, with a 123% increase among those over 60. But only 7% would let an AI book an entire trip on its own, and just 4% fully trust it. The technology is adopted faster than the confidence to delegate decisions to it.
Your hotel website is now a verification tool, not a discovery tool
Markus Busch, writing for hospitality.today, reframed a statistic that has been circulating for weeks: when an AI recommends a hotel, it cites the property's own website in fewer than 10% of cases, based on a Tharro study of 695 unbranded searches across ChatGPT and Google AI Mode. But Busch argues the number describes how the machine selects which hotel to name - it says nothing about the ten minutes that follow.
Travelers arrive at a hotel website not because they found it through the AI, but because they do not trust the AI's answer. Phocuswright puts trust in AI-generated results at roughly 17-20% in the UK, France, and Germany. About 58% of travelers go directly to the hotel site to verify what the AI told them. They typically arrive with a total price in mind - taxes and fees included - because that is what the OTA or AI summary showed them in one screen. The hotel site then asks for dates before revealing anything, with taxes and fees appearing only at steps three and four. Busch's recommendation: publish the real total, not a teaser rate, as the first number visible everywhere it appears.
Marriott CEO Anthony Capuano told a Bank of America investor conference that "the biggest impact you will see on our sector is the distribution landscape" - how travelers discover a hotel, before the booking itself. Marriott feeds Google and OpenAI content on its portfolio to ensure its properties "appear in a way that is both authentic and visible." The statement confirms at the highest executive level that the battle is shifting to who appears in the conversation, before the question of who collects the payment.
Why this matters for hospitality professionals
The Wyndham tactic - placing your own data directly into the AI models rather than waiting to be cited - represents a shift from passive SEO to active GEO. For chain properties with the technical resources, the playbook is now public: feed structured, real-time data to ChatGPT, Claude, and Gemini, monitor how travelers phrase their searches, and adjust accordingly. For independent properties, the immediate priority is narrower but no less urgent. First, check before September 30 whether your Hotel Ads campaigns depend on a Google-sourced rate, and connect a direct feed if they do. Second, ensure the total price on your own site matches what an AI or OTA shows in one screen - because the traveler who arrives to verify will leave if the numbers do not align. Third, the infrastructure is converging: GDS, PMS, and distribution platforms are all being rebuilt for machine-to-machine communication. Your data needs to be structured and accessible to an agent, not just readable by a human. The gap between AI adoption and real impact is not a technology problem - it is a data readiness problem, and the travelers are already using the tools.
Your membership also unlocks:
