AI & automation
EU AI Act and website chatbots: what you must tell visitors
What Article 50 of the EU AI Act means for your website chatbot: who must disclose, from when, wording that works, and what the Commission says falls short.
If your website has an AI chatbot and your visitors are in the EU, people must be told they are talking to an AI system, clearly, at the latest when the conversation starts. That rule comes from Article 50 of the EU AI Act, and it has applied since 2 August 2026. A short first message such as "Hi, I'm an AI assistant for [business name]" plus a visible "AI" label next to the chat window covers most small business cases. A line buried in your terms and conditions does not.
The rest of this article explains who carries the obligation, what the European Commission's July 2026 guidelines count as a good disclosure, the narrow "obvious" exception, and a checklist you can use this week.
What Article 50 actually requires
Article 50 is the transparency part of the AI Act. For chatbots, the relevant rule is paragraph 1: AI systems intended to interact directly with people must be designed so that those people are informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person.
Paragraph 5 adds the "how": the information must be given in a clear and distinguishable manner, at the latest at the time of the first interaction, and it must meet applicable accessibility requirements.
Other parts of Article 50 cover different situations that may also touch a small business:
- AI-generated content. Providers of generative AI must mark synthetic output in a machine-readable way.
- Deepfakes. Deployers who publish AI-generated or manipulated images, audio or video that look real must disclose it.
- Emotion recognition and biometric categorisation. Deployers must tell people the system is operating.
For a typical website chat widget, paragraph 1 is the one that matters.
The dates
The Commission's fact page on the transparency rules is direct: they apply from 2 August 2026. The AI Omnibus, which entered into force on 27 July 2026, moved several high-risk deadlines (Annex III systems now from 2 December 2027, Annex I products from 2 August 2028). It also gave generative AI systems already on the market before 2 August 2026 a grace period until December 2026 for the machine-readable marking duty. That grace period is about marking content. It does not delay the duty to tell people they are chatting with an AI.
So if your chatbot is live and serves EU visitors, the disclosure rule already applies.
Provider or deployer: who is responsible
The AI Act splits responsibility between two roles.
- Provider: the company that develops an AI system, or has it developed, and places it on the market or puts it into service under its own name or trademark.
- Deployer: the business that uses an AI system under its authority in a professional activity.
Article 50(1) is formally a provider obligation: the system must be designed so people are informed. In practice, the line is less tidy for small businesses, and the Commission's guidelines show why. A company that builds a chatbot in-house and puts it into service under its own name for its own use is a provider. A company that takes an existing generative AI system, modifies it (for example with new training data) and puts it into service under its own name becomes the provider of that new system.
Many website chatbots sit in this zone. You pick a platform, feed it your service pages and price list, give it your brand name and avatar, and put it on your site. Whether that makes you a provider depends on the facts, but the safe position is simple: make sure the disclosure appears in your chat, whoever is technically responsible. If your vendor's widget doesn't show it by default, add it yourself in the welcome message. Check your contract too, since it should say who handles AI Act compliance.
What a good disclosure looks like
The Commission published guidelines on Article 50 on 20 July 2026. They are the most practical source available on this question, and they give concrete examples.
Methods the guidelines describe as appropriate
- Text in the interface: a prominent, plain-language label or banner such as "You are interacting with an AI system", and a first-turn greeting from the chatbot that says it is AI. These can be combined with a badge that stays visible during the conversation.
- Placement: disclosures close to the interaction itself, for example near the input field, with simple wording for people with lower digital literacy and for children.
- Voice and phone: an explicit spoken statement at the start, such as "This is an AI-powered assistant", with reminders in longer calls. A tone or sound cue alone is not enough.
- Visual cues: persistent icons or recognisable "AI" symbols that complement the text.
The guidelines also say that a single, prominent notification before the first interaction is likely to be enough in most cases. Riskier contexts, such as conversations with children, older people or people with disabilities, call for reminders.
Methods the guidelines say fall short when used alone
- A disclosure only in your terms and conditions, a URL or documentation.
- Invisible markers such as metadata.
- Vague signals, such as calling the bot a generic "assistant", or human-like personas that may mislead.
- Site-wide statements like "Services on this website use AI", which aren't specific to the chat.
- Technical descriptions like "this system uses LLMs" that don't tell the user they are talking to an AI.
That fourth point catches a lot of small business sites. A footer note about AI does not do the job of a disclosure inside the chat.
Sample wording
Keep it short and put it in the first message, in the language of the page:
- English: "Hi, I'm [Name], an AI assistant for [Business]. I can answer questions about our services and book a call. A person from our team reviews every booking."
- Greek: "Γεια σας, είμαι ένας βοηθός τεχνητής νοημοσύνης (AI)."
- French: "Bonjour, je suis un assistant IA de [Entreprise]."
Add a small "AI" tag next to the bot's name or avatar. If you give the bot a human first name, the AI label matters even more, since a named persona with a human photo is exactly the kind of signal the guidelines call misleading.
The "obvious" exception is narrow
Article 50(1) doesn't apply where it is obvious that the person is dealing with an AI. The guidelines treat this as an exception to be read carefully. The fact that people know chatbots exist does not make any particular interaction obvious. Human-sounding voices, natural writing and a profile photo all make it less obvious.
The Commission gives examples on both sides. Code assistants used only by professional developers can qualify. AI chatbots embedded in online platforms or helpdesks, where users receive replies they may perceive as written by a person, are listed as examples where the obligation applies. A customer-facing chat on a business website falls on that side. Don't rely on the exception.
Penalties
According to the Commission's fact page, breaches of the transparency rules can lead to fines of up to €15 million or 3% of total worldwide annual turnover for companies, with proportionate treatment for SMEs and small mid-caps. National market surveillance authorities enforce them. For a small business the realistic risk is less about the maximum fine and more about a complaint, a regulator's letter and the cost of fixing things in a hurry. A disclosure costs almost nothing to add.
Your chatbot also handles personal data
The AI Act sits on top of the GDPR. Chat conversations often contain names, phone numbers, emails and details about a person's situation. Your privacy notice should mention the chat, say who the chatbot vendor is, how long transcripts are kept and what they are used for. If the bot sends leads to a CRM or books appointments, that flow belongs in your privacy notice too.
Checklist for a website chatbot serving EU visitors
- Open your chat on desktop and on a phone. Does the first thing a visitor sees say it is an AI?
- Add an "AI" label next to the bot's name or avatar, visible for the whole conversation.
- Remove anything that suggests a human, such as a stock photo of a person, unless the AI label is next to it.
- Write the disclosure in every language your site uses.
- If the bot can hand over to a person, say so, and make the switch visible ("You're now chatting with Maria from our team").
- If you use an AI voice agent on the phone, make sure it says it is AI in its first sentence.
- Ask your vendor in writing how its product supports Article 50, and keep the answer.
- Update your privacy notice to cover chat transcripts.
- Recheck after every vendor update or redesign. Widgets change, and disclosures sometimes disappear with them.
This article explains the rules in plain terms and isn't legal advice. If your chatbot gives advice in a regulated field such as health, finance or law, have a lawyer review the set-up.
How PNE handles this
When PNE sets up an AI chat assistant through our AI Growth Systems service, the disclosure is part of the build: the first message says it is an AI, the widget carries a label, and every language version gets its own wording. It is a small detail, and it is far easier to build in on day one than to retrofit after a complaint.
Sources
- Article 50: Transparency obligations for providers and deployers of certain AI systems — EU Artificial Intelligence Act
- Quick facts: transparency rules for AI systems — European Commission
- Guidelines on transparency obligations for providers and deployers of AI systems — European Commission
- Guidelines on the implementation of the transparency obligations under Article 50 (PDF) — European Commission
- AI Act: regulatory framework and timeline — European Commission
- AI Omnibus enters into force — European Commission