An AI Concierge for Your Short-Term Rental: What Is Safe to Automate and What Is Not
published 28 September 2026
Artificial intelligence has entered short-term rental management through the inbox. The promise is simple: an assistant that answers guests at any hour, in any language, without you picking up the phone. That promise holds for part of the conversation and is risky for the rest. This article suggests a practical way to tell the two apart, covers privacy in broad terms and gives you questions to ask before letting a tool talk to your guests.
Automating Answers Is Not Automating Decisions
The most useful distinction is this. A factual answer repeats something you have already decided and written down: the wi-fi password, check-out time, where the bins are. A decision creates something new: agreeing to a late check-out, offering a discount, saying a fault is harmless, promising someone will be there in half an hour.
An AI concierge can do the first kind of work well, as long as it sticks to what you wrote. The second kind should stay with a person, because it involves responsibility, money or safety, and because guests will treat the answer as your commitment.
What Is Safe to Automate
Questions with a fixed answer that you have already written and that do not change from guest to guest:
- Wi-fi network and password
- Check-in and check-out times
- How appliances work: air conditioning, water heater, washing machine
- Where the rubbish goes and how to sort it
- Parking, transport, nearest supermarket and pharmacy
- House rules: quiet hours, smoking, visitors, pets
- Recommendations you left: restaurants, beaches, walks
These are exactly the questions guests ask most, often at times when you are not available. Answering them in seconds, in the guest's language, improves the stay and saves you time.
What Should Always Go to the Host
- Breakdowns and problems in the home: no hot water, a leak, an appliance that will not start, a lock that will not open. The AI can gather details, but should not diagnose or play the problem down.
- Complaints: an unhappy guest needs a person. An automated reply to a complaint tends to make it worse. If a guest asks for the complaints book, they must get it, and you need to know. See our article on the complaints book.
- Money: refunds, discounts, deposits, extra charges, price changes. Any amount an AI promises may be treated as a commitment.
- Booking changes: early arrivals, late departures, extra guests, date changes.
- Emergencies: health, safety, fire, break-ins. The right response is to point to 112 and alert you immediately, not to try to solve it.
- Anything not written down: if you did not provide the information, the AI should not make it up.
The Main Risk: Made-Up Answers
Language models are good at producing plausible text, and that includes plausible, wrong text. In a rental, a made-up answer might be a door code that does not exist, a restaurant that has closed, a rule you never set or a promise of a free late check-out.
There are three practical defences:
- Limit the source. The concierge should answer only from what you wrote about that home and that stay, not from general internet knowledge.
- Admit when it does not know. Faced with a question you have not answered in writing, the right behaviour is to say it will pass it to you, not to improvise.
- Hand over with context. When it passes a conversation on, you should get an alert with what the guest asked, so you can reply without starting over.
Privacy and GDPR in Broad Terms
An AI concierge processes personal data: names, contact details, stay dates, the content of messages. This is not legal advice, but some principles of the General Data Protection Regulation (GDPR) are worth keeping in mind:
- Tell guests what data is processed, why and by whom, including the use of AI.
- Minimise: do not ask for or keep more than necessary, and do not use the concierge to collect sensitive data without need.
- Know where the data goes: which providers are involved, where the data is processed, how long it is kept and whether it is used to train models.
- Have a contract with the provider: anyone processing data on your behalf should have a data processing agreement with you.
There are also AI-specific rules. The EU Artificial Intelligence Act sets transparency obligations, applying from 2 August 2026 under its phased timetable, for systems that interact with people: they should know they are talking to an AI unless that is obvious. If you are unsure about your own case, check with Portugal's data protection authority (CNPD) or a lawyer.
How to Evaluate a Tool
Before letting an AI concierge talk to your guests, ask:
- Where do the answers come from? Only from what I wrote, or from general knowledge too?
- What happens when it does not know? Does it make something up, or pass it to me?
- How am I alerted? Email, WhatsApp, notification? With what context?
- Can I see the conversations? To fix missing information and understand what guests ask.
- Do guests know they are talking to an AI?
- Which languages does it answer in, and how well?
- Where is the data stored and for how long? Is there a data processing agreement?
- What does it cost and how is it charged? Per message, per stay, per month, by commission?
Test it as if you were a guest: ask the easy questions, then the hard ones, then one the guide does not answer. How it handles the last one tells you more than the others.
How ciceroni Helps
ciceroni's concierge answers guests around the clock, on WhatsApp and in each stay's digital guide, in Portuguese, English, French and Spanish, using only what you wrote about the home. When it cannot answer, it does not improvise: it hands the conversation to you and sends you an alert by email or WhatsApp. It is free during launch; after that, pricing will be per stay, with no monthly fee and no commission.
FAQ
Does an AI concierge replace the host?
No. It answers the factual questions that keep coming, such as wi-fi, times or appliances, and leaves breakdowns, complaints, money, booking changes and emergencies to you.
What if the AI gives wrong information?
The risk drops considerably when the concierge only answers from what you wrote and passes on what it does not know. Even so, review conversations and fix any missing or outdated information in the guide.
Do I have to tell guests they are talking to an AI?
EU transparency rules for AI provide that people should know when they are interacting with an AI system, unless that is obvious. It is also a matter of trust: guests should know when they are talking to you and when they are not.