Do You Have to Tell Guests They're Talking to AI? The EU AI Act and Automated Guest Messaging (2026)

Yes, in practice. Since 2 August 2026, Article 50(1) of the EU AI Act requires that people who interact directly with an AI system are informed they are dealing with an AI, from the start of the first interaction, unless that is obvious from the context. If an AI answers your guests on Airbnb, Booking.com, Expedia, WhatsApp or by email, guests in the EU should be told once, in plain words, and a human should remain reachable.

The legal duty sits mainly with the provider of the AI system, the company that builds it, which must design the disclosure in. Hosts who use a ready-made AI co-host are deployers. What matters for you is that the tool you use handles the notice correctly, and that you know what to say if you build your own bot.

This article explains the rules for hosts. It is not legal advice.

What does Article 50 of the AI Act require?

The European Commission's guidance on transparency obligations under Article 50 sets four cumulative conditions for the chatbot duty:

• the system is an AI system;
• it is designed for a genuine two-way exchange;
• the AI communicates directly, not through a human intermediary;
• the other party is a natural person.

Chatbots, AI agents and avatars are the examples the Commission gives. The person must be informed from the start of the first interaction. The only exception is when the AI nature of the exchange is obvious to an average, reasonably well-informed person, and the Commission says that exception is to be read restrictively. A guest writing to "the host" on Airbnb has no reason to assume a machine answers, so the exception rarely applies to guest messaging.

Who is responsible: the software provider or the host?

Article 50(1) is addressed to providers. If you use an AI co-host built by a software company, that company has to make sure guests are informed. If you assemble your own bot, for example with an AI model connected to your inbox through automation tools, you become the provider of that system and the duty is yours.

Deployers have separate duties in Article 50(3) and 50(4). They cover emotion recognition, biometric categorisation and deepfakes, and AI-generated text published to inform the public on matters of public interest. A reply to a guest about the Wi-Fi code is none of these. Your usual obligations still apply: your privacy notice should mention that guest messages may be processed by automated tools, as GDPR requires transparency about processing.

Does a reply you review before sending count?

Usually not. The third condition says the AI must communicate directly, not through a human intermediary. When an AI only drafts a reply and you read, edit and send it yourself, the guest is talking to you. That is the copilot model most messaging tools use.

The duty applies when the AI sends on its own, which is exactly what makes an autonomous AI co-host useful: answers in a minute at 3 a.m., without you approving each message. Autonomy and disclosure go together.

Does it apply to non-EU hosts and to phone lines?

The AI Act applies when the output of an AI system is used in the EU, so a host outside the EU who answers EU guests with an AI is concerned, and EU hosts answering non-EU guests should apply the same notice for consistency. Voice assistants that talk to guests on the phone are interacting directly with people too: the caller should hear that an AI is speaking at the start of the call.

How do you tell guests without sounding robotic?

• Say it once, at the start. One sentence in the first AI reply of a conversation is enough. Repeating it in every message adds noise.
• Keep it warm and short. For example: "Hi Anna, I'm the AI assistant for this home and I reply around the clock. If you'd rather talk to a person, just say so." Adapt it to your tone and the guest's language.
• Offer a human. The notice reassures most when it comes with a way out. Make sure "I want to speak to someone" reaches you.
• Don't hide it in the house rules. A line in your listing or welcome guide is good context, but the notice belongs in the conversation itself.

What are the fines?

Breaches of the transparency obligations can be fined up to €15 million or 3% of worldwide annual turnover, whichever is higher, with proportionality for small companies. National market surveillance authorities enforce the rules. The separate obligation to mark AI-generated content in a machine-readable way (Article 50(2)) is also a provider duty; systems already on the market before 2 August 2026 have until 2 December 2026 to meet it. There is nothing for hosts to do there.

How Nowistay handles it

Nowistay's AI co-host replies on its own, so the notice is built in: since August 2026 it adds a short, one-time disclosure to its first reply for travellers in the EU and EEA, in the guest's language, and every conversation can be taken over by you at any moment. If you want to watch the AI before letting it send, draft mode emails you the reply it would have sent instead of sending it. To see everything the assistant handles, read what the Nowistay AI co-host does, and for the bigger picture, how AI is changing vacation rentals.