Where do hotels already use AI?
Hotels already use AI in five places, and in each one it takes the repetitive part: it answers guests on chat, moves the price of each room night, plans housekeeping and maintenance, drafts the first reply to an event enquiry and writes the marketing. Decisions and the welcome stay with people.
The firmest examples are in BCG’s 2026 study of the «AI-first» hotel. At some hotels, AI pricing has lifted revenue per available room, RevPAR, by more than 15%, according to the analysts at STR. At the Ritz-Carlton San Francisco, an AI system that matches cleaning to check-outs cut the time to prepare a room by 20%. At the Four Seasons Peninsula Papagayo in Costa Rica, AI monitoring cut food waste by roughly half.
| Department | What AI does today | What stays with people |
|---|---|---|
| Guests, chat and phone | Answers routine questions at any hour (parking, times, check-in), in the guest’s language | Complex requests, complaints, the welcome |
| Room rates | Moves the price of every night with demand, competitors and the pace of bookings | Deciding how much risk to take, and when to refuse a rate |
| Housekeeping and maintenance | Orders cleaning around check-outs, flags a fault before it happens | Checking the room, doing the work |
| Events and groups | Drafts the first quote and the reply email | The final price, the call, the site visit |
| Marketing | Writes offers, pages, newsletters, posts and review replies | Choosing what to promote and approving what goes out |
In 2025 Hotel Tech Report asked 402 hotel guests how they feel about AI during a stay: 70% find chatbots helpful for simple questions and still want a person for anything complex. That is the split I would make: the 3 a.m. parking question goes to the machine, everything else to the front desk.
Why now?
Hotels are turning to AI now because staff are scarce and cost more. In the same study, BCG quotes the American Hotel & Lodging Association: in 2025, 65% of North American hotels reported staff shortages. BCG adds that labour costs rose 11.2% in a year.
65%
of North American hotels reported staff shortages in 2025, while labour costs rose 11.2% in a year.
65%: American Hotel & Lodging Association; 11.2%: BCG, «AI-First Hotels», February 2026.
European hotels tell us the same thing in their own words. Between October and November we spoke with 21 hotels and hotel groups, from a single villa to a chain, and in 12 of them marketing is one person with no time. «The marketing department is only me», said the marketing lead of three seasonal hotels. What they need from AI is the work that keeps slipping, starting with the newsletter and the social posts.
Are guests already asking AI about hotels?
Guests already ask ChatGPT about hotels, and we can see it on our own site. Lumina runs a hotel guide for travellers. Between 23 October and 19 November it received 1,725 visits from ChatGPT, counted on our site, Counted the same way on our site, Google Search brought 1,676 visits over the same days, so for our guide ChatGPT is already as big a front door as Google. Most of those visitors had a practical question about a hotel they already had in mind: what a night costs, what the buffet costs, whether outside guests can use the pool.
- 1,725
- visits from ChatGPT to our hotel guide in 28 days
- 1,676
- visits from Google Search over the same days, counted the same way
- 1,386
- AI answers where the guide appears, in four countries
Visits counted on Lumina’s site, 23 October to 19 November 2026; AI answers measured by DataForSEO on 21 November 2026 in the US, the UK, Italy and Germany.
The guide appears in 1,386 answers from AI assistants and is cited as the source in about a third of them. Roughly 4 questions in 10 ask for a price: what a night costs, what the spa costs. Then, on 21 November, we put 52 guest questions to ChatGPT and to Google’s AI Mode ourselves. When a hotel’s own website stated the fact, such as the price of its day spa for non-residents, ChatGPT cited the hotel. When it did not, ChatGPT looked elsewhere, and six times, always for a price, it cited us. The guide also works inside ChatGPT and Claude, as our own app for the assistants, which answers about hotels from checked facts.
So the lesson for a hotel is simple: put your facts on your own site, in plain words, kept up to date. Indicative prices, opening times, rules on children and pets, what outside guests can book. To see what ChatGPT says about your hotel today, our free AI visibility check shows it in a few minutes; for what to publish, read how to get your hotel on ChatGPT and Gemini.
What does AI still get wrong?
AI still gets facts wrong: it writes fluently and invents with the same confidence it uses for the truth. In October we asked a model for the facts of 31 Maldives resorts: of the 77 answers we could check, 38 were wrong. The full test is in what AI gets wrong about a hotel’s facts.
38 of 77
answers wrong about the facts of 31 Maldives resorts, in our first test.
Lumina’s own check, 3 October 2026.
Today our guide is written this way: one AI drafts the page, and a second one, which did not write it, checks every fact against the hotel’s own source. In one November night, across some 40 pages, that check made about 300 corrections: a «from» price that was not the lowest, a review given the wrong date, a future date told as if it had already happened. In a hotel the last check is a person: AI writes the draft, someone at the hotel signs it off.
Why does marketing fall behind?
Marketing falls behind because it is the one job without a slot in anyone’s day. Rates have their system, chat has its own, housekeeping has a head of department; the newsletter and the blog wait for somebody to find an hour. In August we read the websites of 669 of the world’s top-ranked hotels: fewer than one in four kept a blog alive.
That is the job we built Lumina for. It reads your website, reviews, photographs and the area around your hotel, finds the nights that need more guests, and prepares the offer, the page, the email and the ads in your hotel’s voice. It uses your own photographs, never invented images, and nothing goes out without your OK. How it works, and what AI already does well in the other departments, is on our page about AI for hotels.
How should a 30 to 120-room hotel start?
A hotel of 30 to 120 rooms should start with one job, the one that costs it most today, judged on its own numbers. This is where I would begin.
Pick the job that slips most
Night-time chat, room rates or marketing: one only, the one nobody does well today.
Put your facts on your website
Indicative prices, opening times, rules and what outside guests can book, in plain words: it is what the assistants read when a guest asks about you.
Try it for a month against one number
Questions answered without the front desk, RevPAR on quiet nights, direct bookings: one number, before and after.
Leave the last word to a person
AI prepares, someone at your hotel approves. It is the rule that stops the wrong price and the promise you cannot keep.
Ask where the data goes
Where it is hosted, who receives it, whether it trains a model. A clear answer, in writing, before you sign.
What does European law ask of a hotel?
European law asks two practical things of a hotel. The first comes from Article 50 of the AI Act, the EU’s regulation on artificial intelligence: people talking to an AI system must be told so, unless it is obvious, and an image or video made or altered with AI that could pass for real must say so. For a hotel, that means a chat that introduces itself as an automated assistant, not as the receptionist.
The second is about guests’ data, which stays personal data when a machine reads it: before you connect a tool to your property management system or your inbox, ask the questions in the last step above. Lumina, for instance, is hosted in the European Union, and a hotel’s material is used only for that hotel’s work and never trains a model; the details are on our security page.
Questions hotels ask
- Will AI replace hotel staff?
- Not the people guests remember. It takes the repetitive part, the routine question, the Sunday rate change, the newsletter nobody had time to write, while decisions and the welcome stay with your team. With staff this hard to find, the problem today is finding people, not replacing them.
- What does an AI receptionist do?
- It answers routine questions on chat, WhatsApp or the phone, at any hour and in the guest’s language: parking, times, check-in, services. It passes complex requests to the front desk. Under the AI Act it must say it is an automated assistant, unless that is obvious.
- Which AI software do hotels use?
- Usually one per job: a revenue management system for rates, a messaging platform linked to the property management system for guests, tools for housekeeping and maintenance, and for marketing a tool that knows the hotel. Start with one job and judge it on that job’s numbers.
- Can I just use ChatGPT?
- For a draft, yes, if you give it everything first: the hotel, the guests, the tone, the offer. What ChatGPT does not do alone is read your reviews, competitors and local events every week, remember your corrections and prepare the work in your website’s and newsletter’s own template. That difference is on our page about AI for hotels.
- How much does Lumina cost?
- A monthly fee per hotel, set by its number of rooms and agreed on the demo, with no set-up fee; hotels under 25 rooms start free and pay as they go, from €4.50 for 25 credits. Every price is on the pricing page.
Try Lumina on your own hotel

Marco Di CesareFounder of Lumina
Before a demo Lumina reads your website, your reviews and your competitors. In 15 minutes you see what it found and the first work it would do. If you would rather ask me something first, we can talk for 30 minutes.
All of this was read on 21 November 2026, and platforms change their minds often: every claim links back to its page.
Where we read it (4)
- BCG, «AI-First Hotels: Leaner, Faster, Smarter», February 2026
- Hotel Tech Report, AI in Hospitality, with the State of Hotel Guest Technology Report 2025 (402 guests)
- European Commission, AI Act Service Desk, Article 50 (transparency obligations)
- EUR-Lex, Regulation (EU) 2024/1689 on artificial intelligence (AI Act)



