Do I Need to Disclose My AI Chatbot Under the EU AI Act? (Article 50)

• Yes, in almost every case. If your website shows visitors an AI chatbot or assistant, Article 50 of the EU AI Act requires that those visitors be informed they are interacting with an AI, not a human.
• The disclosure must be given at the first interaction, be clear and accessible, and appear in every language you serve.
• The only carve-out is when it is "obvious from the circumstances." That exception is narrow: a conversational support bot does not qualify.
• The obligation applies from 2 August 2026 (Article 50), the same day most of the AI Act's remaining rules take effect.
• Transparency breaches sit in the AI Act's second penalty tier: up to EUR 15 million or 3% of total worldwide annual turnover, whichever is higher (Article 99(4)).
• It applies whether or not you built the bot. If you use a third-party chatbot (Intercom, Tidio, Drift, Crisp, and similar), you are a deployer and you are responsible for making sure the notice is shown on your site.
• Chatbots are only part of Article 50. If you also publish AI-generated text or images, Article 50(4) asks you to label them.
• On WordPress you can add a compliant disclosure in minutes, for free, with the EU AI Act WordPress plugin: it runs locally, needs no account, and no data leaves your site.

The short, direct answer

If your site uses a chatbot, a virtual assistant, an AI voice agent, or any AI system that communicates directly with visitors, you almost certainly need to disclose it. Article 50 of the EU AI Act gives people the right to know when they are dealing with a machine rather than a person. This right does not depend on how "risky" your AI is: even a limited-risk customer-support bot is in scope. The safe default is simple, if a visitor could reasonably think they might be talking to a human, tell them they are not.

This guide walks through exactly who is covered, the one narrow exception, what a compliant notice must contain, what happens if you ignore it, and how to implement it on WordPress in a few minutes.

What Article 50 actually says

Article 50 is the EU AI Act's set of transparency obligations. It has four operative parts, and two of them commonly reach ordinary websites:

• Article 50(1), interacting with AI. AI systems intended to interact directly with natural persons must be designed and used so that those persons are informed they are interacting with an AI system, unless that is obvious from the circumstances.
• Article 50(2), AI-generated content. Providers of systems that generate synthetic audio, image, video, or text must mark that output as artificially generated in a machine-readable way.
• Article 50(3), emotion recognition and biometric categorisation. Deployers of these systems must inform the people exposed to them.
• Article 50(4), deepfakes and AI text. Deployers who publish AI-generated or manipulated image, audio, or video (deepfakes), or AI-generated text on matters of public interest, must disclose that it is artificially generated.

For a typical website with a chatbot, Article 50(1) is the one that bites. If you also publish AI-written articles or AI images, Article 50(4) applies too.

Provider or deployer: who owes the duty?

The AI Act splits responsibility by role (Article 3):

• A provider builds an AI system or puts it on the market under its own name.
• A deployer uses an AI system in the course of its activities.

Article 50(1) is written as a design duty on the provider, the company that makes the chatbot. But in practice, if you run the website, you are the deployer, and you are responsible for how the system is used on your site. The regulator's expectation, and the only safe posture, is that the notice is actually visible to your visitors. So:

• If your chatbot vendor already shows a clear "you are talking to an AI" notice, confirm it is adequate (visible, at first contact, in your languages).
• If it does not, or the notice is weak, you add it. You cannot rely on "the vendor should have done it" as a defence for a notice your visitors never saw.

This is why third-party widgets like Intercom, Tidio, Drift, Crisp, Zendesk, LiveChat, Tawk.to, and HubSpot chat still put the practical duty on you: you chose to deploy them on your site.

Which chatbots and assistants are in scope?

Article 50(1) covers AI systems "intended to interact directly with natural persons." That is broad. It includes:

If your bot uses a language model or any machine-learning component to generate responses, treat it as in scope.

The "obvious from the circumstances" exception

The exception in Article 50(1) is real but narrow. It covers only cases where any reasonable person would already know, without being told, that they are dealing with a machine. Consider:

• Does not qualify: a support bot that greets visitors by name, answers in fluent natural language, and reads like a person. This is precisely the situation Article 50 was written for.
• Might qualify: a widget explicitly and permanently labelled as a "bot" in a context where no human interaction is offered at all, and no reasonable user could be misled.

If you are relying on this exception for a conversational, human-sounding chatbot, you are almost certainly relying on it wrongly. When in doubt, disclose. The cost of a short notice is nothing; the cost of getting the exception wrong is a transparency breach.

What a compliant disclosure must do

The regulation does not dictate exact wording, but a defensible notice meets four practical tests:

1. Timing, at the first interaction. The visitor learns it is an AI before or as the conversation begins, in the bot's opening message or as a visible label on the widget. Not on a separate policy page they must hunt for.
2. Clarity, in plain language. A normal visitor understands it immediately. For example: "You are chatting with an AI assistant. Responses are AI-generated."
3. Accessibility. Real text, not an image of text, so screen readers and assistive technology can read it. Adequate contrast and size.
4. Language coverage. Present in every language your site serves. A German-language visitor should see a German notice.

Good and weak examples

• Good (badge on the widget): "You are chatting with an AI assistant. Responses are AI-generated."
• Good (first bot message): "Hi, I am an AI assistant. I can answer common questions, or connect you to a person."
• Weak: a line buried in your privacy policy. Visitors never see it at the moment of interaction, so it does not meet the timing test.
• Weak: an image containing the words "AI bot" with no alt text. It fails accessibility.

What are the penalties?

The EU AI Act enforces its rules through a tiered penalty regime in Article 99:

So a failure to disclose your AI chatbot falls in the middle tier: up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher. For an SME, national authorities apply proportionality, but the exposure is real, and reputational risk arrives well before any fine. See our full breakdown in EU AI Act penalties and fines explained.

When does this take effect?

Article 50 transparency obligations apply from 2 August 2026. That is the AI Act's main application date for most remaining provisions. Note two nearby dates that also matter for websites:

• 2 February 2025: Article 4 AI-literacy obligations and Article 5 prohibitions already apply.
• 2 August 2026: Article 50 transparency and most high-risk rules apply.

The European Commission has also been developing guidance and a Code of Practice on transparency to help operators implement Article 50 in practice; the direction of travel is toward clear, up-front, machine-readable disclosure, which is exactly what a well-configured notice provides.

How to add the disclosure on WordPress

You do not need a developer. There are two routes:

• By hand: add a visible notice to your chat widget or theme, and maintain it across every language and template yourself. Workable, but easy to get wrong and to let drift.
• With a plugin: a transparency plugin renders the Article 50 notice consistently, in the right place, in multiple languages, in your choice of banner, inline, badge, or modal style.

The Legalithm EU AI Act plugin does this locally, it runs entirely on your site, needs no account, and sends no data anywhere. It also scans your site to detect which AI chatbots are installed and suggests the matching notice, so you switch it on with confidence. Nothing visitor-facing is enabled without your explicit action. For the full walkthrough, see how to add an Article 50 AI disclosure to WordPress.

Do not forget AI-generated content (Article 50(4))

Chatbots are one duty; content is another. If you publish AI-generated text or images, Article 50(4) asks you to label them as artificially generated so readers can tell them apart from human-made material. This is easy to miss because generated content looks identical to human content, there is nothing to auto-detect, so it depends on you flagging it at the point of creation. If you use AI to draft posts or create images, treat labelling as a separate task from the chatbot notice.

Is a rule-based chatbot (no AI) covered? Generally no. Article 50 targets AI systems. A purely scripted decision-tree menu is not an AI system. But if any part of your bot uses a language model or machine learning to generate replies, treat it as in scope. If unsure, disclosing costs you nothing.

My chatbot is from a third party (for example Intercom or Tidio). Am I still responsible? Yes, in practice. You are the deployer. If the vendor's built-in notice is clear and shown at first contact, confirm it and you are fine. If not, you must add the disclosure yourself.

Does the EU AI Act apply to my non-EU website? It can. The Act reaches organisations outside the EU where the AI system's output is used in the EU, or where you target EU users. If EU visitors interact with your chatbot, assume Article 50 applies.

Do I need the visitor's consent to run an AI chatbot? Article 50 is about transparency (informing), not consent. You must tell people it is AI; you do not need their permission under Article 50 to operate the bot. Separate rules (for example data protection under the GDPR) may still apply to the data the bot processes.

What exactly does the notice have to say? There is no mandated wording. It must clearly convey that the person is interacting with an AI, at the first interaction, in plain, accessible language, in the right language. "You are chatting with an AI assistant. Responses are AI-generated" is a solid default.

Does disclosing the chatbot make me fully AI Act compliant? No. It handles one transparency duty. Your risk tier and any duties beyond transparency depend on what your AI actually does. Run the free EU AI Act assessment to map your specific obligations.

Get this handled before 2 August 2026

The deadline is fixed. If you run a WordPress site with a chatbot, the fastest path is to add the disclosure now and confirm the rest of your obligations with the assessment.

• Add the notice (free, local, no account): EU AI Act WordPress plugin
• Check your full obligations: Free EU AI Act assessment
• Understand the rule: Article 50 transparency obligations

This article is general information, not legal advice. For your specific situation, consult a qualified professional.