Ask ChatGPT for "a boutique hotel in Paris with a great pool, good for couples, under 200 euros" and it doesn't read your website to answer. It reads what your guests wrote. The hotel industry spent twenty years learning to win a ranking built on links and ad budgets, and the AI assistants that are replacing that ranking don't trust either. They trust reviews.
That is the argument Ben Jost, TrustYou's CEO and chair of HSMAI Europe's AI Advisory Board, made in the first episode of HSMAI Europe's new weekly podcast. We've been saying a version of it for years. What changed in 2026 is that it stopped being a reputation argument and became a distribution one.
Searches used to be two words: "hotel Paris". Then three: "hotel Paris pool". Now people type into an assistant the way they'd brief a friend, and the query is a sentence with four or five conditions in it. Something on your side has to match that sentence. A room description can't. An amenity list can't. A photo gallery can't. The only text that describes a hotel at the level of detail people now search at is the long tail of what guests said about staying there.
So the data a quality team has been collecting for years for the monthly score is now the data the assistant reads to decide whether to recommend you. As Ben put it, this is "a renaissance of review content", and the practical consequence is blunt: a hotel with thin feedback is a hotel the assistant can't describe. (03:00)
Here is where most hotel marketing goes wrong, and it goes wrong by doing what worked on Google. The instinct is to write for the machine: more pages, more keywords, now more "AI-optimised" copy. Ben's answer to that is the clearest moment in the episode. Describe what you offer, let guests describe it on top of you, and make sure the two agree. "If you're describing 'we have the most amazing breakfast' and then you have a bunch of reviews that say the breakfast wasn't that good, you're going to be down-ranked. You're not going to be found for breakfast at all." (20:55)
The models are good at knowing who wrote what. A hotel's own copy is a claim; a thousand guests are evidence. When the claim and the evidence disagree, the evidence wins, and the claim costs you the query. Content written to game the assistant fails for the same reason, because nobody games a thousand witnesses.
Our position is that this is good news, if you're honest. The hotels that lose under this rule are the ones whose websites were always slightly better than their breakfasts.
None of these need a new platform, and all of them are things we watch our best-performing customers do already.
Collect more than you think you need. Send the survey. Ask on every channel. Volume is what gives the assistant something to match against, and it is also what makes the score statistically honest instead of hostage to one bad week.
Make it visible, including the bad ones. Publish reviews on your own site. Respond to the negative ones instead of burying them; a visible, answered complaint reads as a hotel that pays attention, and the assistant reads the reply too. (03:58)
Put the score where people feel it. The operators Ben rates most highly attach part of management's bonus to it. "You're an 89, your bonus is attached to a 90, and suddenly people care." Then stop reading dashboards and start asking questions: what happened last week, what's the action plan for housekeeping, which comments are pulling the score down. (06:36)
If reviews are now distribution, then the way most hotel groups own them is broken. Quality bought the feedback software, so quality runs it, and marketing, which carries the "how do we get found" target, has nothing to do with it. Ben called this out directly, and it matches what we see in almost every group we work with. The two teams that need the same data make decisions in separate rooms. (16:10)
The fix is organisational and it starts at the top: one view of guest feedback, one set of decisions about it, and a small cross-functional group that owns both the score and the visibility. The hotels that do this will be describable to an assistant. The ones that don't will keep running two strategies for one asset.
There is a version of this story where AI search becomes Google with a new referee, and the biggest budgets win again. Ben doesn't believe it, and neither do we. When you ask a personal assistant for the best hotel in Lisbon, you expect a recommendation made for you. You don't expect whoever paid the most. That expectation is the opening. For the first time in two decades, an independent hotel doesn't have to outspend a platform to be in the conversation with a guest. It has to be good, be described honestly, and be reachable. (22:23)
The catch, which Ben is candid about, is that the industry hasn't yet worked out how to connect into the assistants together. Every brand is building its own door. His estimate of the window is somewhere between 12 and 24 months, and that is the reason the AI Advisory Board and this podcast exist. (23:55)
The reward for getting this right isn't a ranking. It's time. The picture Ben paints of the hotel that has made the shift is a person no longer sitting in front of a screen: the arrival list sorted by VIP with a history of each guest, rooms allocated, the birthday ticket raised, all from two instructions to an assistant that can see the PMS and the guest profiles at once. Hours of typing becomes a conversation, and the people get to go back to doing hospitality. He is clear that this is already possible rather than a forecast, and so are we. (13:06)
Your guests have already written the description the assistants are looking for. The only question is whether you're collecting it, showing it, and living up to it.
They match the traveller's request, which is usually a full sentence with several conditions, against text that describes each hotel in that much detail. Guest reviews are the only source with that detail at scale, so a hotel's review volume, recency and content weigh more than its own descriptions. Where the two disagree, the reviews win.
Writing copy to game the assistant doesn't work, because the models distinguish a hotel's claims from guests' accounts and discount the claims when they conflict. What helps is describing what you offer accurately, collecting enough feedback to be describable, publishing it, and replying to it.
AI search visibility is whether, and how often, an assistant names your property for the kinds of requests your ideal guests make. The practical measure is to run the twenty queries your guests would type, in each major assistant, once a month, and track which properties appear; monitoring tools for this are starting to exist, but a spreadsheet is enough to start. The inputs you control are review volume, review recency, consistency between your copy and your reviews, and visible responses.
Watch the full episode on YouTube. HSMAI Europe Weekly Conversations publishes a new episode every Wednesday.