Do I have to label a chatbot on my hotel website?

Yes, unless it is obvious that the system is an AI. The duty under Article 50(1) falls on the provider of the system. As the operator, you need to make sure the notice appears on your website. The safest approach is for the assistant to introduce itself as an AI at the start of every conversation.

Do AI-written blog articles have to be labelled?

Article 50 requires disclosure for AI-generated text that is published to inform the public on matters of public interest, unless it has been reviewed by people and approved editorially by someone responsible. For ordinary hotel blog articles, this usually does not apply. Even so, it is worth documenting who checks your texts before they are published.

Who checks compliance?

Member States appoint market surveillance authorities. In Germany, the Federal Network Agency (Bundesnetzagentur) has been responsible since 29 July 2026. Austria has not yet designated one, according to the RTR. For the current position, contact the AI service desk (KI-Servicestelle) of the RTR. It serves as a point of contact and information centre on AI and supports the implementation of the AI Act.

Has the Digital Omnibus postponed the labelling duty for chatbots?

No. The Digital Omnibus Regulation (EU) 2026/1744 mainly postponed the rules for high-risk systems to December 2027 and August 2028. Article 50 has applied since 2 August 2026. Only for machine-readable labelling of generative systems that were already on the market before that date does a deadline run, until 2 December 2026. The position is as of October 2026.

Does the notice need to appear in every language of the website?

The regulation requires clear and unambiguous information, and it does not set a specific language rule. In practice, the notice should be in the language in which the guest is served. A German notice on an English-language interface hardly serves its purpose. Plan the translation in from the start, so that every guest sees the notice in the language they use.

EU AI Act Art. 50 for Hotels: Chatbot Labelling Rules

Key takeaways

The transparency obligations under Article 50 have applied since 2 August 2026.
If you offer an AI system that interacts directly with people, it must be designed so that people can tell they are speaking to an AI.
Deepfakes, meaning realistic-looking, artificially generated images, videos or audio, must be disclosed.
The information must be given clearly and recognisably no later than the first interaction.
The Digital Omnibus Regulation of July 2026 did not postpone the chatbot disclosure duty; mainly the rules for high-risk systems were postponed.
This article does not replace a legal review. For your specific case, consult a lawyer.

The regulation in two sentences

The EU AI Act regulates AI systems according to their risk. Most applications in hotels and holiday apartments, such as an assistant for guest questions, do not fall under the strict rules for high-risk systems, but they do fall under the transparency obligations of Article 50 [Q1] [Q3].

What Article 50 specifically requires

Interaction with people: Providers must design AI systems intended for direct interaction with natural persons so that those persons are informed that they are interacting with an AI system. Exception: this is obvious to a reasonably well-informed person from the circumstances (Article 50(1)) [Q2].

Synthetic content: Providers of systems that generate audio, image, video or text must mark the outputs as artificially generated in a machine-readable way. Operators who create deepfakes must disclose that the content is artificially generated or manipulated (Article 50(2) and (4)) [Q2].

Timing and form: The information must be given clearly and unambiguously, and in an accessible way, no later than the time of the first interaction or exposure (Article 50(5)) [Q2].

The European Commission published guidelines and answers to frequently asked questions on the transparency obligations in July 2026 [Q5]. As permissible forms of notice, the RTR lists, among others, written and explicit spoken notices as well as graphic labelling [Q7].

The question of roles is important. A provider is whoever develops an AI system, or has it developed, and places it on the market or puts it into service under its own name; an operator is whoever uses it under its own responsibility [Q1]. A hotel that buys a ready-made chat solution is usually an operator and can have it confirmed how the notice is handled. Anyone who has a bot developed and operates it under their own name may be a provider themselves; this should be checked by a specialist.

What Article 50 does not require: no registration, no certification, and no technical documentation as for high-risk systems.

Which deadlines apply and which are still to come

Regulation (EU) 2024/1689 entered into force on 1 August 2024 and applies in stages; the dates are set out in Article 113 [Q1]. The Digital Omnibus Regulation (EU) 2026/1744, published in the Official Journal on 24 July 2026 and in force since 27 July 2026, has postponed some of them [Q6]. As of 4 October 2026, the position is:

2 February 2025: prohibitions on certain practices and the obligation to ensure AI literacy
2 August 2025: obligations for providers of general-purpose AI models, i.e. the language models behind many chat solutions
2 August 2026: general date of application, including the transparency obligations under Article 50
2 December 2026: deadline for machine-readable marking under Article 50(2) for generative systems that were already on the market before 2 August 2026
2 December 2027 and 2 August 2028: rules for high-risk systems, depending on classification

The Omnibus has not postponed the chatbot notice duty. According to the Commission, a transition period exists only for machine-readable marking under Article 50(2) [Q5] [Q6].

For a hotel, this means the duty already applies. It falls first on the provider of the system; as the operator, you make sure the notice reaches your guests [Q2] [Q5]. Check whether the notice is in the right place, including for systems that no one any longer thinks of as AI, such as a translation module for guest messages.

What this means for your property

Chat on the website: introduce yourself as an AI at the start, with a permanent notice in the chat window
Replies via WhatsApp or email: labelling in the first message, notice in the signature
Phone assistant: announcement at the start of the call
AI-generated photos of the property: label them, or better, use real photos
Texts checked by people: generally no labelling needed; document the review

Bild: Ein antiker Messingschlüssel liegt auf einem rustikalen Holzschlüsselbrett auf einem Holztisch in der Hotelhalle, Morgenlicht fällt durch große Fenster mit Bergblick herein.

What must appear on your site

For the chat on the website, we recommend four elements:

a notice in the chat window, visible before the first input and in the language of the interface
a first reply from the system that repeats this, without lecturing the guest
a section in the privacy policy: which inputs are processed, for how long, and with which service provider
a visible route to a person, such as a phone number, a callback request or handover to reception

A sentence in the terms of use, or a bot that evades the question of whether a human is replying, is unlikely to be enough; the information must be clear and unambiguous [Q2]. Also avoid first names and portrait pictures that make the assistant seem like a person.

Typical cases from practice

The booking chat of a software provider: Even if the chat comes from a service provider, it is used on your website with guests. Ask how the labelling is handled and document the answer.

Automated replies in the email inbox: If an AI writes replies that are sent without review, the message should include a notice. If your team checks every reply before it is sent, the situation is different; document that.

Images for social media: An AI image that looks like a real photo of your property but shows rooms that do not exist is not only a transparency issue; it can mislead guests. Use real photos of your property.

A property with 24 rooms and three channels: how many notices are needed

A worked example with assumed figures: a property with 24 rooms replies automatically through three channels (website chat, messenger and an automatic first reply by email), each in German and English.

AI notice per channel and language: three times two, i.e. six text modules
Route to a person per channel: three modules
Section in the privacy policy per language: two
Total: eleven places that must be maintained

With a fourth channel, such as an answering machine, there are fourteen. More important than the number: the notices sit in different systems, and without a list one of them goes out of date, usually in the channel that was added last. How automated replies in messenger apps are set up is shown on the page on guest communication via WhatsApp.

Who is responsible in Austria and Germany

Each Member State must designate at least one market surveillance authority to monitor compliance (Article 70) [Q1]. In Austria, the first point of contact for questions is the AI service desk (KI-Servicestelle) of the RTR. It is based on § 20c KOG (Communications Authority Act) and § 194a TKG 2021 (Telecommunications Act 2021), serves as a point of contact and information centre on AI, and supports the implementation of the AI Act [Q4]. It does not issue an approval for your chatbot, as Article 50 does not provide for one.

According to the RTR, Austria has not yet designated a market surveillance authority for the AI Act (position October 2026). For the current status, ask the AI service desk [Q9]. In Germany, the Federal Network Agency (Bundesnetzagentur) has been responsible since the end of July 2026 under § 2 KI-MIG (German AI Market Surveillance Act) [Q8]. Data protection remains separate: the processing of guest inputs continues to be governed by the GDPR (DSGVO). Clarify both together, so that the notice and your privacy policy fit together.

Checklist

Inventory: Where do you use AI today, including in tools from your service providers?
Labelling: Every assistant introduces itself as an AI.
Transparency page: A page on your website explains use, purpose and contact.
Handover to people: Guests can reach a person at any time.
Data protection: Data processing agreement, retention periods, information in the privacy policy.
Training: Your team knows what the assistant does and what it does not do. Article 4 has required AI literacy measures since February 2025; since the revision of July 2026, this means supporting the AI literacy of staff without having to demonstrate any particular level [Q6].
Log: A sheet with channel, system, provider and date of the last check. After every update, check that the notice is still visible.

Bild: Frühstücksraum eines Boutiquehotels mit gedeckten Holztischen, Servicetheke mit Espressomaschine, natürlichem Licht durch große Fenster und Blick auf bewaldete Berge.

How we handle it

Our AI concierge comes with labelling, a transparency page and handover to your team built in from the start. What this means technically is set out under legally compliant AI assistants.

EU AI Act Article 50: Chatbot Disclosure Rules for Hotels