AI recommends the hotel it knows best, not the best one

BCG says optimize for AI search. Its own data shows why that mostly helps the OTAs

In January, Boston Consulting Group surveyed more than 9,000 shoppers in nine countries. Its advice to brands was simple: people use AI to shop now, so make sure AI mentions you. Spend on showing up in the answer, the way you once spent on showing up in search.

Read the survey closely and the advice falls apart. What people want from AI isn't a list. It's an answer they can stop second-guessing. And the hotels AI feels surest about are the ones the OTAs have described for twenty years.

People let AI decide now

Search helped travelers make up their own minds. AI makes up their minds for them.

The survey is clear about why people like it. They didn't say AI is faster. They said it makes them feel confident about the final choice. More than 60% said they trust what it tells them (BCG, January 2026). And one finding matters more than the rest: in BCG's interviews, people said AI helps them work out what they want in the first place.

That part is new. A search engine waits for you to know what you're looking for. AI tells you. Someone types "somewhere warm in October, good food, not too touristy." The old way, that traveler opened ten tabs and decided for themselves. Now the AI reads the request and hands back a place to stay, already chosen. The traveler didn't pick it off a list. The AI picked it, and the traveler said yes. Once the choice is made inside the chat, being easy to find doesn't help much, because nobody is browsing. There's no list to climb. There's one answer, and either you're in it or you're not.

The booking might still happen on your own site or on an OTA. But that's just the payment. The choice was already made — in the chat, by the AI, in the moment a traveler used to spend comparing tabs.

A confident answer is one name

Here's the part the "just show up in AI" pitch skips over.

When AI recommends a hotel, it isn't judging which one is best. It's going on how much it already knows about each one. The more a place is described — clearly, often, the same way, across lots of sites — the surer AI is when it names it. Quality doesn't come into it. Coverage does.

That coverage is exactly what the big OTAs built. Years of listings, reviews by the thousand, rates and photos kept current everywhere travelers look. AI learned about hotels from that web, so those are the hotels it trusts most.

Picture a traveler asking for a hotel near a convention center. AI will name the big chain it has read about ten thousand times before it risks naming the better-placed boutique it has seen once. It didn't compare the two and prefer the chain. It just couldn't be sure about the small one.

You can watch this happen. Ask ChatGPT about hotels in your own town and see which names it reaches for first — usually the chains and the big boxes, the properties with the deepest pile of reviews behind them. The better independent down the street often doesn't come up until you name it yourself.

So the test quietly changed. It used to be: which hotel fits best. Now it's: which hotel can AI name without hesitating. A boutique can be the best room in town and still be the one AI won't commit to — not because it's worse, but because the internet has said less about it.

Why "show up in AI" is the wrong goal

Put those two findings together and the advice gives out.

If the choice is made inside the chat, and AI picks the name it's surest about, then making your hotel easier for AI to read just enters you in a contest you can't win that way. You're paying to be one of the options in a place that no longer shows options. It gives one answer, and that answer leans toward the hotels described first and most.

And the usual fix makes it worse. How do you get AI to know you better? Distribute more. List on more channels. Collect more reviews. But most of that lands on the OTAs' pages, in the OTAs' words. You do the work of getting described, and the OTA gets the credit, because the describing happened on its site. Twenty years of OTA listings quietly taught the whole internet to talk about hotels in the OTA's voice. That's the voice AI trusts now.

None of this means ignore AI. A hotel AI can't read at all won't get mentioned, and that is fatal. Being readable is the floor. But the floor isn't an advantage. It just keeps you in the room.

What actually helps is harder, and it isn't another feed. It's becoming a hotel AI can name on its own — one with a clear, specific story in your own words, things only you can say about your place, so AI doesn't need a stack of OTA pages to feel sure. That comes down to who you are and what you publish yourself, not how many channels you plug into. The hotel that wins this has something specific to say and says it plainly: the only rooms inside the old town walls, a kitchen the chef actually runs, a view the photos can't fake. Put that on your own site, in plain words, again and again, and AI has something to repeat that no OTA owns.

BCG is right that AI is where people decide now. The advice that follows is backwards. The AI answer isn't a billboard you buy your way onto. It's a verdict — and the verdict is mostly settled before the traveler types a word.

Read also: Hotels are building the AI discovery layer. OTAs are harvesting it.

by Markus Busch, Editor and Publisher of 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 →