When a guest asks AI where to stay, would your hotel make the answer?
New data from Barcelona and Cloudbeds shows independent hotels are 8x less visible in AI travel recommendations, with OTAs supplying 55% of AI-cited sources versus 14% for hotel websites.
Somewhere right now, a guest is asking ChatGPT, Gemini, or Google's AI Mode where to stay in your city. If your hotel is not in the answer, there is no lost booking to trace, no drop in traffic to explain, and no bad review to respond to. The guest simply never hears your name.
That is the quiet change AI has brought to hotel distribution. AI is not just a new marketing channel. It is becoming the layer that decides which hotels guests even consider - before a website visit, before metasearch, before an OTA page.
And at the moment, that layer favors chains and OTAs by a wide margin.
What the new data actually shows
This autumn, Carlo Del Mistro and the team at Stiplo analysed 659 accommodations in Barcelona across 27,360 AI travel answers, from four AI assistants in three languages, with every question asked 20 times because AI answers vary from run to run. The study, presented at Skift's Data + AI Summit in London, found that independent hotels were eight times more likely than chain hotels to be completely unknown to AI.
About one in five hotels was never recommended at all, in any language, by any AI. About 20% of hotels captured more than 80% of all AI mentions.
Yet the same study contains the most important nuance for independents: once an AI knew an independent hotel, it recommended it about as often as a chain. The problem is not the product. It is discoverability. The AI has simply never heard of you.
A second dataset explains why. A Cloudbeds citation analysis of 5,720 citations in AI hotel recommendations found that OTAs accounted for 55.3% of cited sources, while property and brand websites accounted for only 13.6%. In the Barcelona study, hotel websites supplied only about 7% of what AI said about a property. Around 68% came from guides and listicles.
In other words, the website and booking engine a hotel has invested in for years is a minority source for what AI tells a future guest about that hotel. AI answers from the sources it already trusts - OTA listings, review platforms, and editorial guides - precisely the sources chains have systematically maintained for years, and many independents have left to intermediaries.
Discovery is hard. Booking is harder.
On October 1, Google AI Mode began completing hotel bookings in the U.S. through ten major chains and OTAs - the same day Google ended the price-feed mandate that had given independent hotels free ad visibility alongside OTA results. Hotels without a live price feed or Hotel Ads connectivity lost presence at exactly the moment AI Mode became a transaction channel, not just a research interface.
A smaller U.S. study published on October 16 makes the booking effect concrete. Researchers asked ChatGPT, Google AI Mode, and Gemini about Austin hotels 330 times, separating discovery prompts (“best boutique hotels”) from booking prompts (“show hotels I can book now”). When the prompt turned to booking, most independents lost 40% to 73% of their visibility. Chain hotels rose from 7% of discovery answers to 53% of booking answers, while Booking.com and Expedia rose from 3–5% to 22–24%.
Independent tests of live agent booking point in the same direction. When agents can navigate one hotel to checkout and fail at another, they do not retry - they route the traveler to a competitor.
This is not yet about volume. It is about memory.
Skeptics are right on one point: AI booking volume is still small. Booking Holdings has said less than 1% of its room nights come from AI, and Bernstein estimates around 1.6% at Marriott.
That is exactly why now matters. Shortlists form long before transactions do. A substantial share of AI hotel recommendations comes from what a model already “remembers” from its training and trusted sources. That memory is being built in 2026, while few hotels are paying attention.
Independent hotels have seen this pattern before. Those who ignored Google in the late 2000s and OTAs in the early 2010s spent the next decade buying back visibility at 15% to 25% commission. AI threatens to repeat that cycle - only earlier in the journey, and less visibly.
What actually moves AI visibility - and what doesn't
The Barcelona data is also useful for what it did not find. The study reported no measurable independent effect from the tactics now being sold as “AI optimization” - llms.txt files, crawler settings, FAQ markup, and schema packages on their own.
What did move visibility was more familiar, and more operational:
Google review volume - the strongest single driver in the study.
Backlinks and existing Google rankings - authority AI systems already use.
Presence in the guides and listicles AI actually reads - the source of most AI hotel descriptions.
Clear, answer-shaped content - plain-language answers to real guest questions about parking, accessibility, quiet rooms, distances, check-in and late arrival.
Consistent, bookable data - accurate rates, availability and property facts across the direct site, OTA listings and feeds, so AI does not default to the most consistent third-party repository.
That leads to a practical sequence for an independent hotel this quarter:
First, test yourself in AI. Ask ChatGPT, Gemini, Perplexity, Grok, Meta AI and Google AI Mode the questions your guests actually ask, several times each, in your main source-market languages. Note which hotels appear. That is your new competitive set.
Second, make reviews a weekly KPI, not a reputation afterthought - built into checkout, post-stay messages and the front-desk routine.
Third, pitch the guides. Identify the local listicles and travel articles AI keeps citing for your destination, and earn a place in them.
Fourth, fix your price-feed and booking connectivity. If your rates are not in the feed, AI cannot show them - an OTA's rates will fill the gap.
Fifth, add AI tools last. Guest messaging, voice agents and booking assistants amplify a hotel guests can already find. They do not fix invisibility.
The question for your team this week
Do not start with “What is our AI strategy?”
Start with one measurable question: When a guest asks AI where to stay in our city, is our hotel in the answer - and if it is, can AI book it directly?
If nobody in the building knows, that is the answer. And unlike the last distribution shift, this one can still be caught early - but only by hotels that measure their AI visibility now, while the shortlists are still being formed.
Comments
Comments for this content
0 comments available