What does an AI chatbot for hotels do, and when does it pay?
An AI chatbot for hotels is an assistant that answers guests in writing: in the chat window on your website, on WhatsApp, sometimes on Instagram or in Booking.com messages. It does three jobs. It answers routine questions, it takes guests who want to book to your booking engine with dates and party size already filled in, and it hands the front desk whatever it cannot handle. You will also see it sold as an AI concierge for hotels or an AI receptionist; the job is the same.
It pays in one particular case: when the same questions arrive while nobody can answer straight away. In the evening, at weekends, on heavy arrival days, in languages nobody at the desk speaks. In 2025 Hotel Tech Report surveyed 402 guests about technology in hotels, and the result draws the line neatly: 70% find chatbots helpful for simple questions, but still want a person for complicated ones.
- 402
- guests surveyed about technology in hotels
- 70%
- find chatbots helpful for simple questions, and want a person for complicated ones
- 58%
- say AI improves booking and the stay
- 65%
- want hotel technology more advanced than what they have at home
Hotel Tech Report, State of Hotel Guest Technology Report 2025.
I would put one in a hotel that gets plenty of repeated questions in several languages and has a booking engine to send guests to. In a 30-room hotel whose guests are mostly local and whose phone rarely rings, the same money goes further on a website that answers for itself, with its facts written down properly. The hotel chatbot is one of the five places where AI already works in hotels, which we covered in how hotels use AI.
What does it do well, and where does it go wrong?
A hotel chatbot is good at questions with one answer, written down somewhere. What time breakfast starts, whether parking is charged, whether dogs are allowed, what the airport transfer costs, whether there is a double free on Friday. It answers at once, in the guest’s language, and the conversation stays on record, so the next morning you know what people asked. HiJiffy, for example, shows the team the questions its assistant could not answer, with a link to add what is missing.
Providers promise big numbers. HiJiffy says its assistant resolves up to 90% of conversations on its own; Canary tells the story of a hotel that automated 82% of its guest enquiries. These are the providers’ figures about their own customers: the one that matters to you is your hotel’s, after the first month.
It goes wrong in three places. The first is the complicated request: a complaint, two connecting rooms for a family travelling with a grandfather in a wheelchair, a quote for a wedding. That needs a person, and a good chatbot notices and hands over the conversation with everything the guest has already written. The second is the fact it cannot find, or finds in the wrong place. In October we had a model fill in the fact sheets of 31 Maldives resorts from what the web says about them: of the 77 answers we could check, 38 were wrong, most often because it trusted the wrong source among several.
38 of 77
answers wrong about the facts of 31 Maldives resorts, when a model filled in their fact sheets from the web.
Lumina’s check, 3 October 2026.
The third is tone. An assistant that passes itself off as the receptionist, or pushes an upgrade on someone who only asked when check-out is, costs more trust than it earns. So I would put the same question to every provider: where do your answers come from, and what do you do when you cannot find one? HiJiffy’s answer is that its assistant uses only content the hotel has approved and, when it is unsure, asks the guest for more detail or passes the conversation to the team. That is the right answer, and a demo with your guests’ real questions is where you check it.
Which hotel chatbots are there, and what do they cost?
Of the eight hotel chatbots I compared, only one publishes the price of its chatbot: HiJiffy. The others quote on request, after a demo, by rooms or by the number of enquiries you get each month. On 21 November we put 20 questions about AI for hotels to ChatGPT, Gemini and Google’s AI Mode: of 80 answers, HiJiffy appeared in 28, Canary in 23 and Asksuite in 15. The table holds what each one says on its own website.
| Chatbot | Channels | Connects to | Languages | Suits | Public price |
|---|---|---|---|---|---|
| HiJiffy | Website; social media, booking sites and email from the Pro plan; WhatsApp as an add-on to Pro and with campaigns in Premium | Booking engine and CRM from Pro, property system in Premium | 3 in Basic, 130+ from Pro | Single hotels to groups; the price depends on the number of rooms | From €99 (see the chart) |
| Asksuite | Website, WhatsApp, Instagram, Facebook, email | Property system, booking engine, CRM | 50+ | Hotels that get many requests for quotes | On request |
| Canary Technologies | SMS, WhatsApp, LINE, WeChat | The hotel’s systems, through its integrations | 100+ | Independents and chains: it says over 20,000 hoteliers use it | On request |
| Quinta, formerly Quicktext (Velma) | Website chat, WhatsApp, Messenger, Instagram, WeChat | 50+ property systems and 100+ booking engines | 38 | Hotels that also want someone to look after their data | Velma on request; the data service starts at $599 a month |
| Chatlyn | WhatsApp, website chat, Instagram, Messenger, Booking.com, email | 13+ property systems connected directly | The guest’s language; website widget in 18 | Hotels and groups; data in Frankfurt, EU-only AI on request | On request, per room; 5 inboxes included per property |
| Otello AI (Zucchetti) | Website chat, email, WhatsApp, phone with an AI voice | Property system and CRS; built in with simpleBooking, Vertical Booking, BookingExpert and Suite 5 | 130+ in chat | Italian hotels on Zucchetti systems; typical set-up in 3 days | Tailored, by enquiries a month |
| Gen.io (Iperbooking) | Website chat, quotes by email | Iperbooking’s booking engine, the most common property systems | Whatever the guest speaks | Italian hotels on Iperbooking’s booking engine | On request, with a free trial |
| Amenitiz (Ami) | Booking.com and Airbnb messages; WhatsApp, email and Expedia on the higher plan | Inside the Amenitiz property system | The guest’s language | Small properties: the cases on its site run from 7 to 33 rooms | By rooms, on request |
HiJiffy’s price list also shows where the limits sit. The cheapest plan covers the website only, 25 topics and 3 languages: for a 30 to 120-room hotel with international guests, the real comparison starts at Pro.
Starting prices. The price depends on the number of rooms; set-up extra, €99 to €599 per 5 properties. HiJiffy’s public price list, 22 November 2026.
I would choose from the systems you already have. On a Zucchetti property system or booking engine I would look at Otello, on Iperbooking at Gen.io, on Amenitiz at the assistant already built in. If WhatsApp is the channel that matters and you want pre-arrival messages tied to your property system, I would compare Chatlyn, HiJiffy’s Premium plan and Canary. From every one of them I would ask for the price in writing, in three lines: set-up, monthly fee and usage. Chatlyn, for instance, charges WhatsApp and SMS use on top of the monthly fee.
Where does a chatbot get its answers?
A chatbot answers from what you give it: your website, your documents, the rates in your booking engine. Chatlyn spells it out: a crawler rereads the hotel’s website on a schedule, picking up policies, FAQs and amenities, and fills in the property’s fact sheet from it. Quinta’s Velma works from 4,200 data points per hotel, from breakfast to meeting rooms. A fact that is written nowhere is a fact the chatbot does not know: it passes the question to the front desk or, at worst, invents an answer.
ChatGPT reads the same facts when a guest asks it about your hotel, and we see that every day on our own hotel guide.
1,725
visits from ChatGPT to our hotel guide in 28 days, against 1,676 from Google search.
Visits counted on Lumina’s website, 23 October to 19 November 2026.
We measured the rest in our study of how hotels show up in ChatGPT and AI search.
When a hotel’s own website stated the fact, ChatGPT cited the hotel; when it did not, it looked elsewhere, as we showed in how hotels use AI. People also use our guide inside the assistants: it is our app for ChatGPT and Claude, which answers questions about hotels from checked facts. If you are wondering where the website chatbot ends and ChatGPT begins, we covered it in ChatGPT app or hotel chatbot.
So the work that pays most comes before the chatbot: writing on your website, plainly and up to date, indicative prices, opening times, rules on children and pets, what non-residents can use. The questions your chatbot cannot answer in its first month are your list of pages to write. What to publish so that ChatGPT finds you as well is in how to get your hotel on ChatGPT.
We do not sell a guest chatbot. Lumina does a hotel’s marketing (offers, pages, newsletters, replies to reviews) and keeps the facts on your website up to date, the ones any chatbot and ChatGPT read before they answer. How it works is on our page about AI for hotels.
What does the law ask of a hotel chatbot?
European law asks two things of a hotel chatbot. 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. So the chatbot introduces itself as an automated assistant, with a name that does not pretend to be the receptionist. The same article asks the provider to mark AI-generated content in a machine-readable way; that is its job, and it is worth asking whether it does.
The second comes from the GDPR. Conversations hold names, phone numbers, dates of stay, sometimes an allergy. The provider processes that data on your behalf, and Article 28 of the regulation requires a contract binding it to you; among other things, it may not pass the data to another provider without your written permission. In practice, ask for three answers in writing: where the data is hosted, who else receives it, and whether it is used to train a model.
Then there is WhatsApp. On one of our calls an Italian hotel had its guests’ phone numbers but not their consent to send them offers. Replying to someone who writes to you and sending a campaign to everyone are two different things: before you switch campaigns on, ask the provider how it collects that consent.
How do you choose one and launch it in 30 days?
A hotel can choose and switch on a chatbot in 30 days if it starts from its guests’ real questions and takes them into the demo. This is how I would do it.
Week one: collect 30 real questions
From email, from WhatsApp, from the desk: whoever is on reception notes them for seven days, with the language and the time they came in.
Write the answers on your website
Every missing answer becomes a sentence or a page on your site: the chatbot will read it there, and so will ChatGPT.
Two demos with the same questions
Have two providers answer your 30 questions live. Ask for the price in three lines and for the data contract.
Switch on the website first, then WhatsApp
Website chat goes live in a few days; for WhatsApp, Meta has to verify your business, and HiJiffy talks of a few weeks.
Week four: read the conversations again
Read the ones passed to reception and the ones left unanswered, fix the website and decide whether to keep it.
What should you measure in the first month?
In the first month I would measure four numbers, all of them readable in the chatbot’s console or in your booking engine. The first is the share of conversations closed without the front desk: if it stays low, I would look first for facts missing from the website. The second is bookings that started in the chat, which your booking engine can see if the link carries one of its own parameters. The third is the list of unanswered questions, which is next week’s work. The fourth is group and event enquiries collected in the evening and at weekends, with a name and an email: without a chatbot they would have waited until Monday.
One number I would leave out: the total number of conversations. It grows by itself as soon as the widget opens on every page, and it does not tell you whether a guest got an answer.
Questions hotels ask
- What are the best AI chatbots for hotels?
- The right one depends on the systems you already run. Widely used names include HiJiffy, Asksuite, Canary, Quinta’s Velma (formerly Quicktext) and Chatlyn; in Italy, Otello AI from Zucchetti and Gen.io from Iperbooking, plus Ami inside the Amenitiz property system. I would start with the one that connects to your booking engine and property management system, and test it in a demo with your guests’ real questions.
- What does an AI receptionist do?
- It answers in writing, and sometimes on the phone, the questions that have a single answer: opening times, services, parking, availability, price. It sends the booking link with dates and guests already filled in and hands complaints and complicated requests to the front desk. Under the AI Act it must introduce itself as an automated assistant unless that is obvious.
- How much does an AI chatbot for a hotel cost?
- The clearest public price list is HiJiffy’s: from €99 a month, billed yearly, for the website alone; from €159 with social media, booking sites and 130+ languages; from €319 with the property system and WhatsApp, plus a set-up fee. The other providers quote on request, by rooms or by enquiries a month. Always ask for three separate lines: set-up, monthly fee and usage, such as WhatsApp messages.
- What is the difference between ChatGPT and a hotel chatbot?
- ChatGPT is the assistant guests use on their own, before they choose a hotel, and it answers from what it finds on the web. A hotel chatbot sits on your website and WhatsApp, talks only about your hotel and can connect to your rates. Both read the same facts: the ones written on your website.
- Is there a free AI chatbot for hotels?
- None of the eight providers on this page publishes a free-forever plan; Gen.io offers a trial. Free general chatbots will answer questions, but they do not know your rates and do not take guests to your booking engine, and for a hotel that is the part that pays.
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 22 November 2026, and platforms change their minds often: every claim links back to its page.
Where we read it (15)
- Hotel Tech Report, AI in Hospitality, with the State of Hotel Guest Technology Report 2025 (402 guests)
- HiJiffy, plans and pricing
- Quinta (formerly Quicktext), pricing
- Quinta, Velma hotel chatbot
- Chatlyn, pricing
- Canary Technologies, pricing
- Canary Technologies, AI Guest Messaging
- Asksuite
- Zucchetti, Otello AI
- Otello AI, pricing (in Italian)
- Iperbooking, Gen.io hotel chatbot (in Italian)
- Amenitiz, Ami AI chatbot (in Italian)
- European Commission, AI Act Service Desk, Article 50 (transparency obligations)
- EUR-Lex, Regulation (EU) 2024/1689 on artificial intelligence (AI Act)
- EUR-Lex, Regulation (EU) 2016/679 (GDPR), Article 28



