The summer ChatGPT picks the hotel: why 7 in 10 Italian hotels are invisible to AI

One guest in three now asks the AI first

The numbers landed within months of each other. The Italian trade press, citing Adyen's Hospitality & Travel Report, reports that one in three Italian travellers used AI tools to plan a trip in 2025, up 77% year over year, a share in line with Adyen's own global polling of 40,000 consumers across 27 countries. Bain & Company projects that by the end of 2026 more than 60% of users could rely on AI assistants to choose where to stay.

A parallel analysis reported by Sky TG24 backs this up: nearly 29% of Italian tourists used AI to plan and book their 2025 holidays; among under-35s usage passes 55%, and in Northern Italy, led by Milan and Bologna, more than four in ten travellers use it. Among younger travellers this has already tipped into majority behaviour. Data presented at the GatewAI conference in Milan (Turismi.AI with Guida Viaggi, Repower and Bluvacanze) show that 59% of Italian Gen Z and Millennials plan trips with AI systems, 41% trust AI recommendations more than their friends' advice, and 60% use AI to choose restaurants.

The reason is concrete: more than three in four users report frustration in the search-and-booking phase, and the assistant removes that friction by turning a hunt through hundreds of options into a direct request. In Italy, "AI assistant" almost always means one thing. According to We Are Social's Digital 2026 report, covered by ANSA, ChatGPT carries 84% of the web traffic generated by AI platforms in Italy, far ahead of Microsoft Copilot (8.55%), Perplexity (4.23%) and Google Gemini (2.08%). If ChatGPT doesn't know your property, most of the AI channel doesn't know it either.

7 in 10 properties never come up

Now the other half of the picture. RevenYou, an Italian revenue-management firm, puts it bluntly: 70% of Italian hotels are invisible to ChatGPT, Gemini and Perplexity. They aren't ranked low. They are absent from the pool entirely. When a guest asks for "a family-friendly agriturismo near Siena", the assistant builds its shortlist without them.

The most rigorous study on this is Lighthouse's, analysed by Hotel Manager Lab: 4,545 distinct prompts across 9 destinations, 5 traveller profiles and 101 phrasing variations produced 49,707 hotel mentions, concentrated on just 2,721 unique properties. City-level coverage is brutal: around 10% of properties get cited in Tokyo, 13% in Paris, and in Park City two-thirds of hotels are never named at all.

And the visible few take nearly everything. The top hotel in Paris alone captures 3% of all market mentions, and the top 100 hotels worldwide account for over 13% of total mentions. In the US, Marriott takes over a quarter of chain-hotel mentions and the top three groups reach 54%. A study of Edinburgh's luxury hotels measured the same pattern at city scale: the top 5 of 19 monitored hotels captured 65% of all AI mentions, leaving the other 14 to split the remaining 35%. For a market like Italy's, built on independents, agriturismi and family-run B&Bs, this is the worst possible dynamic.

Perfect reviews won't save you

Every hotelier's instinct is: I have a 9.2 on Booking, the AI will find me eventually. The data say otherwise. Lighthouse found a weak correlation between ChatGPT visibility and Booking.com review scores: two properties with identical guest satisfaction can meet opposite fates, one recommended, the other never named. What decides it is how often the property has already been mentioned across the web; the quality of the stay barely registers. We've seen the same thing outside travel: you can rank #1 on Google and still be invisible to AI.

There is also a mechanism working against independents: OTA "consensus bias". RevPAR Genius argues that Booking.com and Expedia republish the same structured description of your property across hundreds of affiliate domains, and that AI models read the repetition as consensus. It is a framing rather than a measured finding, but it fits the Lighthouse pattern, where repeated mentions are what drive visibility.

In practice, the AI's picture of your property is assembled mostly from what the OTAs publish.

In Italy, the startup dAIrect made the trade press precisely because it exists to make hotels visible to ChatGPT, with VOIhotels (Alpitour group) among its early adopters. The good news is you can do most of the work yourself.

What you actually do, before peak season

1. Measure where you stand. Ask ChatGPT, Gemini and Perplexity the questions your guests would ask: "hotel with parking in central Verona", "romantic B&B in Ortigia", or something as specific as "agriturismo near Siena that takes dogs and has a pool shallow enough for a toddler". Vary the phrasing, and ask in both Italian and English. International guests planning an Italy trip ask in their own language. If you never come up, now you know. Most brands are losing AI citations without ever noticing.

2. Tell your story outside the OTAs. You fight consensus bias with consistent presence: clear, aligned descriptions on your own site, on Google Business, in local guides, on tourism-consortium and pro loco sites. The AI cites what is already cited. Get mentioned by third-party sources that write about your area, beyond the booking portals.

3. Make your site machine-readable. Schema.org structured data for the property (type, location, amenities, price range) and pages that answer real questions ("we are a 10-minute walk from the station", "breakfast included") instead of brochure poetry. And don't slam the door on AI crawlers: blocking all AI bots costs you visibility rather than protecting it.

4. Monitor over time, not once. Assistant answers shift week to week. What you need to know is when you enter the answer set and when you drop out, and who gets recommended in your place.

Run a mention scan on your brand

Questions hoteliers actually ask