Hyatt Builds an AI Discovery Tool. I Tried It. Didn't Like It.
There's a particular kind of technology launch that gets applauded before anyone actually uses it. Hyatt's new AI-powered discovery tool is one of those.
The press release writes itself. AI. Personalisation. The future of travel search.
Then you try it.
The promise vs the product
The tool invites you to search by "location, amenities, pricing points, trip styles or experiences." That's a genuinely interesting brief. Imagine typing "walking tour in London" and getting hotel recommendations near great walking routes, maybe with concierge packages or local partnerships baked in. That would be useful. That would be AI doing something a traditional filter can't.
Instead, here's what actually happens. You type "walking tour in London." The tool surfaces some suggestions. You click on the London Theatre Walking Tour. You hit "find hotels." And you get... London hotels. Just London hotels. Generic, unfiltered, completely disconnected from the walking tour you just selected. You could have typed "London" and received the same results.
It gets stranger. The map helpfully includes hotels in Cambridge, Brighton, and near Hastings. For a London walking tour query. Whatever model is running behind this, it appears to have a loose relationship with geography.
Historic hotels that aren't historic
Try "historic hotels in Spain." A reasonable query with clear intent.
The tool returns every Hyatt property in Spain. Including the Hyatt Regency Madrid Residences, which, whatever its merits, is not what anyone pictures when they search for historic.
The AI isn't filtering by the meaning of your request. It's just surfacing inventory and hoping the word "hotel" in your query is enough to call it a match.
This is the gap between AI as a concept and AI as a product. The concept is sound. The product doesn't deliver on any of its promises. But I think the more interesting question isn't whether Hyatt built the tool badly. It's whether the tool was worth building at all.
The question nobody is asking
Even if the tool worked perfectly, there's a more fundamental problem that nobody seems to be addressing.
Why would a consumer use a confined AI tool on a hotel brand's website when they already have unrestricted AI assistants on their phone?
Think about this from the customer's perspective. They're planning a trip. They open ChatGPT, or Gemini, or Claude. They ask about walking tours in London or historic hotels in Spain. They get recommendations across every brand, every independent, every price point. The AI draws on the entire internet to give genuinely useful, contextual answers. Why would anyone voluntarily narrow their options to a single brand's inventory at the point in the journey where they're still deciding? My honest read is that for most travellers, the answer is never.
Nobody starts their trip planning inside a brand's ecosystem. They start in the open web, in an AI chat, in a group message asking friends for recommendations. By the time they reach Hyatt.com, they've already decided they're interested in Hyatt. They don't need discovery at that point. They need booking.
Where hotels should actually be investing
The opportunity isn't in building worse versions of tools consumers already have. It's in making sure your properties show up when people use the good versions.
When someone asks ChatGPT for "boutique hotels in Barcelona with rooftop pools," is your hotel in that answer? When Gemini is planning a weekend itinerary for a couple celebrating an anniversary, does it know about your spa package? When Claude is comparing five hotels for a business traveller, does it have your meeting room specs and your cancellation policy?
This is the real AI strategy for hotels. Not building walled gardens that nobody will enter, but ensuring your data, your differentiators, your story are structured and available for the AI platforms that customers are already using.
Meet them where they are. Don't ask them to come to you for a worse experience.
What good actually looks like
There are genuine applications for AI in hospitality, and I think they all share the same characteristic: the hotel holds an information advantage that the open internet doesn't. A concierge layer that knows the property, the guest's preferences from previous stays, and the local area in ways no general model can replicate. Pricing intelligence that responds to demand signals in real time. Operational tooling that surfaces guest needs before they're articulated. These work precisely because the hotel knows things the internet doesn't, and AI amplifies that advantage rather than trying to compete against it.
A discovery tool pitched at undecided travellers is not one of those problems. You're competing against the entire internet's knowledge with a subset of your own inventory. That's not a technology challenge. That's a strategy problem.
The test is simple
Before building any AI-powered customer-facing tool, ask one question.
Does this give the customer a better experience than what they already have access to?
If you're honest about it, the answer for Hyatt's discovery tool is clearly no. And that means you're building technology for the press release, not for the guest.
My view is that the smarter investment is in how your information is structured, findable, and machine-readable, so that when a traveller asks an AI assistant about your destination, your property is part of the conversation. That's not as exciting as launching an AI discovery tool. But it's what actually works.