Most of the numbers written about AI and hotels measure the same thing: how many people are using AI to plan a trip. That figure tells a hotel nothing about whether it is the hotel being named. So we publish both, and we measure the second ourselves.
How many people use an AI assistant?
44 percent of US adults say they use ChatGPT, up from 34 percent a year earlier, and half of US adults now use an AI chatbot of some kind.
Pew Research Center, Americans and AI 2026, published 2026-06-17, read 2026-09-16.
ChatGPT has more than 900 million weekly active users and over 50 million subscribers.
AI Overviews from Google now has over 2.5 billion monthly active users.
The Gemini app has surpassed 900 million monthly active users, up from 400 million a year earlier.
Bing monthly active users reached 1 billion for the first time.
Microsoft reports more than 150 million monthly active users across its family of Copilot products, covering information work, coding, security, science, health and consumer.
This is the ceiling, not the demand. Everyone counted here could ask about a hotel, and most of them never will.
What do people actually use AI chatbots for?
OpenAI's own research on ChatGPT usage finds that three-quarters of conversations are for practical guidance, seeking information or writing, split into 49 percent asking, 40 percent doing and 11 percent expressing.
OpenAI, How people are using ChatGPT, published 2025-09-15, read 2026-09-21.
Anthropic's own data shows the ten most common tasks on Claude.ai made up 19 percent of all traffic in February, down from 24 percent the previous November, so what people ask it keeps spreading wider.
Anthropic, Anthropic Economic Index, published 2026-03-24, read 2026-09-21.
About half of all jobs have had at least a quarter of their tasks performed using Claude at some point, according to Anthropic's own research.
Anthropic, Anthropic Economic Index, published 2026-03-24, read 2026-09-21.
Most of that traffic has nothing to do with travel. A hotel competing for a line inside an AI answer is competing against everyday tasks the assistant already owns, not against other hotels.
Does anyone click through when an AI summary appears?
Pew's own tracking shows 58 percent of people who shared their browsing data ran a Google search that produced an AI summary, and they clicked a traditional result far less often, 8 percent of visits against 15 percent for people who never saw one.
Six in ten US adults say they read the AI summaries search engines show them, and three in ten say they do not, per Pew's own survey.
Pew Research Center, Americans and AI 2026, published 2026-06-17, read 2026-09-21.
An AI summary sits above the list of links, so a searcher who trusts it rarely scrolls down to what used to be page one. Getting named inside that summary matters more than ranking beneath it.
How many travellers plan a trip with one?
91 percent of Americans say they plan to travel in 2026, and 49 percent want to travel more than they did the year before, according to a Harris Poll survey run for Marriott Bonvoy.
Marriott Bonvoy, The Harris Poll, published 2025-12-09, read 2026-09-21.
8 percent of travellers in the US and the UK rely on an AI chatbot or agent when they plan a trip.
Expedia Group, The AI Trust Gap, published 2026-04-14, read 2026-09-16.
53 percent of travellers are comfortable letting AI suggest where to go, and 48 percent say it saves them time and finds places they would have missed.
Expedia Group, The AI Trust Gap, published 2026-04-14, read 2026-09-16.
Nearly 70 percent of travellers said they prefer booking with a trusted travel brand over an AI chatbot or agent.
Expedia Group, The AI Trust Gap, published 2026-04-14, read 2026-09-21.
Almost everyone plans to travel this year, and only a sliver of them let an AI help plan it. The distance between being comfortable with a suggestion and actually relying on one is the story inside that sliver, because comfort measures willingness rather than behaviour.
Can a traveller book inside the answer?
Booking.com and Expedia were among the seven pilot partners when apps launched inside ChatGPT, so a traveller sees live options without leaving the conversation.
ChatGPT shoppers can buy directly from Etsy sellers in chat, and more than a million Shopify merchants were coming soon at launch.
Marriott launched Ask Bonvoy, a conversational search experience that helps Bonvoy members explore its portfolio of about 10,000 properties in 146 countries and territories.
Marriott International, published 2026-06-16, read 2026-09-21.
Google rolled out hotel booking inside AI Mode with ten launch partners: Booking.com, Choice Hotels International, Expedia, Hilton, Hotels.com, IHG Hotels and Resorts, Marriott International, Priceline, Trip.com and Wyndham Hotels and Resorts.
Google’s AI Mode is moving the same way, toward booking a hotel without leaving the conversation. Whichever hotels get named first when that happens will matter more to a small property than to a group, because the large chains and agencies already have the visibility a listing buys.
What does AI read when it picks a hotel?
One audit of 1,357 grounding citations across 156 hotel queries in Tokyo measured which sources Gemini leans on when it recommends a place to stay.
arXiv preprint, not peer reviewed, published 2026-03-01, read 2026-09-16.
The audit behind that figure is a preprint, so it has not been peer reviewed, and we read it as a first look rather than a settled account of what Gemini cites.
What do the numbers say actually swings an AI’s hotel pick?
An algorithm audit of AI hotel picks found that a top guest rating raises the odds of being recommended by 31.6 percentage points and a high price lowers them by 30.0, while a hotel's position in the list on its own is worth roughly $12 a night.
arXiv preprint, not peer reviewed, published 2026-06-15, read 2026-09-21.
Asked to price a hotel room's view, GPT-4o put the premium at HK$115.66, well under the HK$771 a separate human study found travellers actually pay.
arXiv preprint, not peer reviewed, published 2026-02-10, read 2026-09-21.
Neither study has been peer reviewed yet, so read both as an early look rather than a settled account. Even so, the pattern matches what our own record shows: a strong rating and a fair price carry real weight, and where a hotel happens to sit in the list can move the answer on its own.
Which hotels do the engines actually name?
ChatGPT, Claude and Google AI named 41 different hotels across Basel, Geneva and Zurich in September 2026, from 46 names given.
The y8y AI Visibility Index, September 2026. 46 names, 3 engines, 3 cities.
5 of the hotels named in September 2026 were named by more than one answer.
The y8y AI Visibility Index, September 2026. 41 businesses, 3 engines, 3 cities.
We put real guest questions to each engine, in a browser, and transcribe the answer in full. Nothing here is estimated, and nothing is written from memory about a business: the names, the lines and the prices are quoted from the run.
How big are the hotel brands competing for the same answer?
Hilton's global portfolio includes 9,158 hotels with more than 1.35 million rooms in 143 countries and territories, and more than 260 million people belong to its Hilton Honors programme.
Hilton Worldwide Holdings Inc., published 2026-07-28, read 2026-09-21.
IHG's hotel business spans 6,963 open hotels with more than one million rooms in over 100 countries, and its loyalty programme, IHG One Rewards, counts more than 160 million members.
InterContinental Hotels Group PLC, published 2026-02-17, read 2026-09-21.
Accor's ALL loyalty programme and H World's H Rewards now count a combined 430 million members after the two companies expanded their partnership.
Booking Holdings' room nights reached 325 million in the second quarter of 2026, up 5.3 percent from a year earlier, and 662 million across the first half of the year.
Booking Holdings Inc., Form 10-Q, published 2026-08, read 2026-09-21.
Expedia Group's gross bookings reached $35.53 billion in the first quarter of 2026, up 13 percent from a year earlier.
Expedia Group, Inc., Form 10-Q, published 2026-05, read 2026-09-21.
None of that scale decided which hotel our own record actually named. The independents and boutique properties that get picked are competing against portfolios and loyalty programs built to be found everywhere, and still losing that particular fight.
What does the hotel AI recommends cost?
The hotel AI recommends in Basel costs a median of CHF 200 a night, across the 11 recommendations that carried a price in September 2026.
The y8y AI Visibility Index, September 2026. 11 names, Google AI, Basel.
The hotel AI recommends in Geneva costs a median of CHF 217 a night, across the 9 recommendations that carried a price in September 2026.
The y8y AI Visibility Index, September 2026. 9 names, Google AI, Geneva.
The hotel AI recommends in Zurich costs a median of CHF 236 a night, across the 10 recommendations that carried a price in September 2026.
The y8y AI Visibility Index, September 2026. 10 names, Google AI, Zurich.
A price is published only where enough of the recommendations carried one, because a median over a couple of hotels describes those hotels and nothing beyond them.
Do the engines agree with each other?
The agreement rate needs one question answered by two engines in the same city. The first city to have one publishes it here.
Operators tend to check one engine and conclude something about all of them. Once this figure qualifies, it will be the reason that check is worth less than it looks.
Does ranking first on Google mean being named?
It does not, and we published the study that shows it. Across a handful of hotels’ own Google data, the pages named in AI answers were tied to what each property was already known for, and several of them ranked outside the first page in ordinary search. The numbers, the method and the limits are on that page rather than repeated here.
See the study: The page that ranks first is not the page the AI names.
How often do the names change?
The churn needs two consecutive months of the same question in the same city. The first comparison publishes here after the second run.
A hotel named in one month is not named for good. This is the figure an operator should watch, because it says how quickly a place that appears can disappear, and it needs a second consecutive month before it can publish.
What do the named hotels have in common?
What the named hotels have in common is recorded per hotel as each run lands. The first pass publishes here.
We record what each named property had on its own site and profiles at the time it was named, so the answer comes from the hotels the engines chose rather than from a theory about them.
The short-stay picture
Guests spent 144.3 million nights in EU short-stay rentals booked through Airbnb, Booking or Expedia in the first quarter of 2026, 9.7 percent more than a year earlier.
Across 2025, guests booked 952 million nights in the EU through those three platforms, 11.4 percent more than in 2024.
Airbnb Nights and Seats Booked reached 305 million for the six months ended 30 June 2026, up 10 percent from 277 million a year earlier.
Short-stay rentals get the same treatment as hotels once the property manager runs are recorded, and those figures will appear here beside these.
The y8y AI Visibility Index
The measured figures on this page are one month of the index. Each is computed from the record when the site is built, never typed, so a figure here and the record it links to cannot disagree. The independent share is the headline. The median price, the agreement rate and the churn publish only when enough of a month exists to support them, and say so when they do not.
Every month’s record is a permanent page with its own table and a CSV. The figures are also available as JSON, and each card above can be embedded with its link back.
How we run it
Each question is put to each engine in a browser and the answer is transcribed in full, with the order the hotels came in, the line the engine gave and the price it quoted. A business is one business however an engine spells it. An international chain is matched on the brand as a whole word, so a hotel whose name merely contains one is not counted as one.
- September 2026 The first record: Basel, Geneva and Zurich, with ChatGPT also in Geneva and Claude also in Zurich, and the first pass of external figures.
- September 2026 Added the sections a browser verification pass unlocked: what people actually use a chatbot for, whether anyone clicks past an AI summary, what drives an AI's hotel pick, and how big the chains and platforms already are.
Questions this study answers
How many travellers plan a trip with an AI assistant?
Fewer than the coverage suggests. Expedia Group surveyed travellers across the US, the UK and India and found that only a small share rely on an AI chatbot or agent when planning a trip, while a much larger share say they are comfortable letting one suggest options. The gap between being comfortable and relying on it is the whole story, and the figures above carry the exact share, the survey's method and the date we read them.
Which hotels does AI actually recommend?
We record it every month and publish the list. In the Swiss cities we run, the engines name a small set of independents and boutique properties far more often than the international chains, and the figure on this page is computed from that record rather than estimated.
Do the AI engines agree with each other?
That is one of the figures we are building toward. It needs one question answered by two engines in the same city before it can publish, and this page says in words when it has not qualified yet rather than printing a guess.