Reference guide · 11 chapters
AI assistants now sit between travellers and hotels, and they do not behave like a search engine. This guide collects our research on what they recommend, which sources they draw on, and why the same question can produce different hotels on different engines. Measuring your own standing is covered in the measuring guide; the practical work of improving it is covered in the GEO guide.
Opinions about AI search are cheap; measurements are not. We run real traveler questions through the AI engines at scale and publish what comes back. Two findings anchor everything else: search visibility only partially carries over to AI answers (a hotel in Google's top 3 is named by AI about 51% of the time — and in roughly two-thirds of cases where AI names a hotel with its own working website, that website is absent from Google's organic results for the same search), and the hotel's own website is cited as the source in under 10% of recommendations. The engines also disagree with each other far more than most hoteliers assume: ChatGPT and Google's AI Mode overlap on just 21% of the hotels they name.
The studies below go deep on how each engine builds its answers, which sources it trusts, what the language of an AI recommendation reveals, and how recommendations differ by traveler persona.
Chapter 1 · 10 min read
We ran 700 traveller questions through Google AI Mode, ChatGPT and Google Search across three markets to see who AI recommends, and who owns the answer.
Read the chapter →Chapter 2 · 4 min read
We asked ChatGPT and Google's AI Mode the same 695 hotel questions. They agreed on one hotel in five. The overlap data, and why AI visibility is two games.
Read the chapter →Chapter 3 · 4 min read
We traced 13,680 links inside real AI hotel answers. Hotels' own websites got under 9%. Who owns the click layer, and how to fight for it.
Read the chapter →Chapter 4 · 4 min read
Everyone assumes AI concentrates hotel visibility into a few winners. Data on 3,380 recommended hotels says the opposite: wide, shallow, open to independents.
Read the chapter →Chapter 5 · 4 min read
Same Google rank, different AI payoff: top-3 hotels get named by AI 56% of the time in the Algarve, 22% in Mallorca. Market-by-market AI visibility data.
Read the chapter →Chapter 6 · 6 min read
20,000+ AI prompts across 30 major European cities: hotels capture 47% of booking authority, OTAs 53%, despite appearing in just 16% of mentions. Why.
Read the chapter →Chapter 7 · 13 min read
We analysed 1.3 million words of AI hotel recommendations across 8 European markets: how ChatGPT and Perplexity describe luxury hotels, and what drives it.
Read the chapter →Chapter 8 · 13 min read
We analysed 71,401 AI hotel recommendations across 30 European cities: the language AI uses for Couple, Family, Business, Ultra Luxury and Wellness personas.
Read the chapter →Chapter 9 · 13 min read
We analysed 720 AI answers across ChatGPT and Google AI Mode for Paphos, Limassol and Ayia Napa. Hotels get recommended; they just don't own the answer.
Read the chapter →Chapter 10 · 11 min read
We asked ChatGPT and Google AI Mode 520 holiday questions in five languages. Three countries dominate, and AI's favourite destination is 1.9× its real demand.
Read the chapter →Chapter 11 · 4 min read
When an AI assistant recommends a hotel it credits a source. We looked at whose domains those sources belong to. Hotels are rarely the answer.
Read the chapter →Measure this for your hotel
Tharro measures how visible your hotel is on ChatGPT, Google AI Mode and Perplexity — continuously, against your competitors.
Track your AI Visibility →Five minutes to set up. One market free for 14 days.