85% of Hotels Never Appear in AI Answers: Destination AI Forum 2026

Roughly 85 percent of hotels never show up when a traveler asks an AI assistant where to stay. That figure, from research by Chris Anderson of Cornell and Nick Slavin of Curacity, anchored the distribution talk at Destination AI Forum 2026. The third edition of the event ran September 29-30 at the National Housing Center in Washington, D.C. It sold out, drawing more than 500 attendees from over 280 companies.

The mood has shifted. Anil Aggarwal of Milestone put it plainly on day two: nobody asks anymore whether to do this. They ask how.

Where hotels stand

Josiah McKenzie of Hospitality Daily opened with The State of AI in the Hotel Industry study, based on answers from more than 100 hospitality leaders. Nine in ten said AI already speeds up routine work. Nobody plans to spend less on AI in 2027. The most common negative feeling was being overwhelmed.

Harman Singh Narula of Canary Technologies shared a number from tens of millions of guest messages. Only about a third are questions, such as what time breakfast starts. The rest are requests, such as late checkout or a broken shower. Most current tools answer questions only. Narula expects AI that executes the requests itself, by arranging the checkout or reporting the repair, to grow quickly over the next 6 to 18 months.

What is working, with numbers

The proven cases came from the back office, and the operators named their savings precisely:

• Food and beverage. Nancy Wolff of Columbia Hospitality described an AI tool that read an operator's invoices and found a salmon price change costing about $2,900 a month.
• Procurement. HHM Hotels and Folio found about $370,000 a year in potential savings, using software that suggests a cheaper approved product during ordering.
• Hiring. Commonwealth Hotels and Ponte screened more than 2,400 applicants by phone in three months, saving 685 manager hours across eight hotels.

Wolff argued the best projects start not with IT or a big system, but with hotel staff who face the problem daily. Dan Kornick, CIO of Loews Hotels, applied a blunt test: if an AI feature costs more than the people doing the job today, he does not buy it.

What does AI visibility mean for 2027?

AI visibility (whether an assistant names your hotel in a recommendation) depends less on what a hotel says on its own site than on how often trusted outside sources, such as travel articles and reviews, mention it. In the Cornell-Curacity study, hotels mentioned most often by those sources were recommended about three times as often as those mentioned least.

The Kahala, a luxury resort in Hawaii, worked on outside mentions and is now recommended five times as often as the least-mentioned hotel in its competitive set. Accuracy matters too. Aggarwal of Milestone shared a case where AI tools claimed a New York boutique hotel offered room service years after it stopped. One new website page fixed that within 30 days.

Greg Land of AWS told hotels to rewrite their 2027 distribution strategy now, or AI platforms will simply connect to existing intermediaries. Actual bookings through AI assistants remain rare. Wyndham and Mobi.AI, builders of one of the first hotel booking tools inside an AI assistant, admitted adoption is still low and the value so far is mostly in what they learned.

What is holding hotels back

Only 8 percent of study respondents blamed the technology's limits. Three barriers recurred:

• Data. Only 42 percent of leaders trust their data readiness. Mike Baldinger of Evention cited a transaction code meaning different things at two hotels in the same group.
• The frontline. 84 percent said their own work is improving, but only 15 percent see gains for frontline teams. Shane O'Flaherty of Microsoft called frontline workers the forgotten people of this technology.
• Verification. Kelly McGuire of Kasa found an owner report where AI credited a revenue gain to the wrong sales channel. Her rule: if AI saves ten hours, spend five checking the result.

Training topped the answers. Keryn McNamara, CIO of Aimbridge Hospitality, said what keeps her awake is no longer cybersecurity, but whether her associates are fluent in AI.

Debates, showcases and the brands

Day one featured the first hospitality AI start-up competition, with six companies pitching. StayFull, an AI-native property management system for small independents, won both the judges' and the people's choice.

A new AI Showcase format paired suppliers with operators for seven minutes each. Accor and Azira ran a targeted campaign for the relaunched Sofito New York aimed at brand-agnostic travelers. DiamondRock Hospitality and PepperMill connected systems at Cavallo Point, giving department heads a daily suggestion on which shift to swap to avoid overtime.

The Great AI Debate put three PMS leaders on stage: Adam Harris of Cloudbeds, Jacob Messina of Stayntouch and Richard Valtr of Mews, moderated by Sloan Dean of AI Hospitality Group. Harris argued for one unified platform, since value gets lost in the gaps between systems. Messina preferred a focused PMS with strong partners, so hotels can replace what fails. Valtr took a middle path: core apps built in, an open system for the rest. The audience voted Messina the winner.

Esther Hertzfeld of Questex led a guest-experience panel on the theme that AI works best when guests do not notice it. Branigan Mulcahy of Virdee told of a luxury hotel that hosted a one million dollar wedding; a year later, the couple returned for their anniversary via an OTA, and unlinked guest profiles prompted the front desk to ask if it was their first stay.

Day two closed with the big brand panel, led by Stuart Greif of Forbes Travel Guide, with leaders from Hilton, Hyatt, Marriott, Wyndham, Choice and IHG. Pat Nestor of Hyatt said the focus has moved from adoption to absorption: are people getting value from tools they already have? Lindsay Liszewski of Choice said the challenge is choosing what to do first. Nitin Sood of Marriott called the approach tech-enabled, human-centered.

Jim Harding, CEO of Otonoma, closed day one with the question of whether we should, not just could, citing an Amazon recommendation engine rushed to market and pulled after it hurt customers. His advice, test more than feels necessary, will likely define which of the forum's promised 2027 gains actually arrive.