AI for hotels: booking, guest requests and automated reviews

The AI for hotel It serves mainly in three moments: requests that arrive before booking, questions from guests during the stay and reviews after departure. In each, an assistant connected to email, WhatsApp and portal messages responds immediately and in the guest's language, starting from the official information of the facility, and passes to the reception what a person requires.

There are, however, things that AI must not do in place of management software. In Italy, an accommodation facility has precise obligations: the communication of the people staying at the police station within twenty-four hours of arrival, the monthly statistical survey on customers, the tourist tax, the national identification code in the ads. These are exact flows that belong to the PMS and to those who respond to them, and the first rule of an AI project in the hotel is not to mix them with conversations.

Where AI enters a hotel's day, and where not

The clearest way to decide what to entrust to the assistant is to follow the host's path and separate conversations from fulfillments.

The industry already has data on which AI works well. According to the ISTAT report Enterprises and ICT 2025, CRM is adopted by 54.9% of housing businesses, and in the housing and catering sector 34.6% of businesses make at least 1% of revenues online. The use of AI in the same sector is 11.7%, below 16.4% of the average of companies with at least 10 employees: many facilities have master records and reservations in order and do not use them yet to respond better.

Requests before booking: answer immediately without inventing rooms

Those who write in the evening to ask if there is parking or if breakfast is included are comparing multiple facilities. The assistant responds at that time, drawing on a knowledge base with official information: check-in and check-out times, services, rules on animals, directions to arrive, cancellation policies. The method of building it is in AI for the business knowledge base.

On free rooms and prices the rule is strict. A channel manager connects the management software of the rooms to the portals and updates rates, availability and reservations on all channels, so a sold room disappears from the others, as explained by the Booking.com Partner Guide. If the assistant reads from there, he can give true answers; if he is not connected, he refers to the site booking engine or to the reception, without confirming anything.

The channel manager also matters for messages. Booking.com allows channel managers and business management systems that integrate it to use its messaging API, so conversations with guests arrive in one place (page about connectivity providers). Ask your provider if they support it: an assistant who sees email, WhatsApp and portal messages in the same queue works better than three separate tools. For the WhatsApp channel the demo WhatsApp chatbots for businesses, in multiple languages shows operation, while the choice of the right language for each guest is explained in Multi-language automatic messages.

From August 2, 2026 also applies the transparency provided by the Article 50 of the AI Act: the guest must know that he is writing with an AI system, unless it is obvious. A sentence at the beginning of the conversation and an easy way to ask a person is enough, and they improve trust more than they reduce it.

Alloggiati Web, ISTAT, tourist tax and CIN: what remains with the business management software

The State Police Web Accommodation Portal He remembers it in the opening: the managers of the accommodation facilities must communicate to the Police Headquarters the details of the people staying within 24 hours after arrival, exclusively through the Accommodation Service, and on arrival itself if the stay lasts less than 24 hours (Article 109 of the TULPS and ministerial decree of 7 January 2013). It is a stream that the PMS handles with exact data. The assistant may remind the guest of the documents to be carried or collected in advance the missing data, but the identification methods must be verified with the Police Headquarters and the document collection channel must be decided with the privacy consultant.

The ISTAT survey on the movement of customers in receptive establishments is monthly and mandatory for Article 7 of Legislative Decree 322/1989; the collection at the facilities is organized by the statistical offices of Regions and Autonomous Provinces, as explained by the ISTAT Circular on Detection. Arrivals and attendance figures come from the business management software: no model must estimate or reconstruct them.

On the tourist tax, the Art. 4 of Legislative Decree 23/2011 makes the manager responsible for the payment, with right of recourse on the guest, of the electronic declaration by 30 June of the following year and the other fulfillments of the municipal regulation. That is why the assistant's answers on rates and exemptions must come from the regulations of your Municipality, word for word. How you ask for payment before arrival is told in automatic payment of the tourist tax.

The national identification code instead concerns the texts. The Art. 13-ter of Decree-Law 145/2023 asks the hotel and non-hotel facilities to display the CIN outside and to indicate it in every advertisement everywhere published and communicated. If you use AI to write descriptions, posts, or offers, the CIN goes into the starting patterns.

Reviews: what changes with the law 34/2026

From April 7, 2026, Chapter IV of Law 34/2026 regulates online reviews of restaurants and tourist facilities, including accommodation facilities.Article 19 considers a review lawful when submitted within 30 days of the stay by someone who actually used the service, consistent with the establishment and not resulting from discounts, benefits or other advantages offered by the provider or its intermediaries; it is presumed authentic if accompanied by tax documentation and ceases to be lawful two years after publication. The establishment’s legal representative, or their delegate, may request its removal when it does not meet these requirements. Article 20 prohibits buying and selling reviews, and the rules do not apply to those already published before entry into force.

For a hotel this changes especially the automation of the review request.

On the answers the AI remains useful: it prepares drafts that take up the points cited by the guest, in his tone and in his language, that the director rereads before publishing, especially when the review is negative. The demo Respond to Google reviews with AI shows the flow. AI can also group reviews by theme (cleaning, noise, breakfast) and report what they look like without a corresponding stay, leaving the decision on a possible report to who represents the structure.

Three structures, three starting points

Examples are scenarios built to show choices, not real customers.

A three star of 40 rooms in an art city. International clientele, reception unstaffed at night. The first step brings email, WhatsApp and portal messages into a single queue through the channel manager; the assistant answers practical questions in Italian, English and German and prepares draft replies to reviews, which the manager approves every morning.

A seasonal lakeside hotel with 25 rooms. Staff changing every season and spikes in demand in the summer. The assistant serves the staff first of all: a knowledge base on procedures, services and rules of the Municipality's tourist tax, which can be consulted by those who have just arrived, and only after direct responses to guests.

A four-room B&B run by a couple. Requests arrive in the evening on WhatsApp. The assistant answers on check-in, parking and breakfast, sends the direct booking link without ever confirming dates, and uses text templates that already contain the CIN. The rest of the evening passes to the couple with the conversation summarized.

Mistakes That Cost a Accommodation Facility

• Confirmed rooms without PMS. An unrelated assistant who "sees" non-existent availability produces overbooking and negative reviews.
• A discount in exchange for the review. With the law 34/2026 this is enough to make even a sincere review illegal.
• Announcements without CIN. Quickly generated texts for portals and social forget the code: put it in the templates.
• Invented tax exemptions. The chatbot that promises an exemption not provided for by the municipal regulation charges the difference to the structure.
• IDs collected where it happens. Photos of identity cards scattered between personal chats and email boxes are a security and privacy issue, even before organizational.
• Answers published without rereading. A wrong public response to a negative review remains visible to all future guests.

The boundary between automatic response and passing on to the person is the same as any project of AI for customer care, with an extra attention to tones: in the hotel the guest remembers how he was treated.

Where to start in your facility

• Collect the questions that the reception receives each week and write down the official answers, including the rules of your Municipality's tourist tax.
• Choose the first channel: often WhatsApp or the email box, then the messages of the portals via the channel manager.
• Ask the PMS provider or channel manager what data the assistant can read and how.
• Prepare advert and offer templates with the CIN already included.
• Set the review request without incentives, sent after departure only to those who stayed.
• Write the sentence that tells the guest that he is talking to an assistant and how to reach out to a person.

For specific use cases of accommodation facilities there is the page AI for tourism and hospitality; for the bases of the channel most used by Italian and foreign guests, WhatsApp chatbots for businesses.

If you want an assistant to answer your guests by reading from your PMS and hand them over to the front desk when needed, look at how we design them AI agents and tell us about your structure.