AI chatbot disclosure examples: wording that actually satisfies Article 50

Article 50 tells you that you must inform users they are interacting with an AI system — it does not hand you the sentence. This guide gives concrete wording you can adapt, and explains why some popular formulations fall short.

What the notice legally has to achieve

Article 50(1) requires that natural persons interacting with an AI system are informed that they are interacting with an AI system, unless this is obvious to a reasonably well-informed person. Article 50(5) adds the delivery requirements: the information must be provided in a clear and distinguishable manner at the latest at the time of the first interaction, and conform to accessibility requirements. In practice that means:

• It says the user is talking to AI — plainly, not euphemistically.
• It appears before or at the first exchange, not buried in terms.
• It is visually distinguishable (not grey-on-grey fine print) and accessible to screen readers.

Wording that works

For a support chat widget, shown as the first message or a pinned header:

• “You’re chatting with an AI assistant. You can ask for a human at any time.”
• “This is an automated AI chat. Answers are generated by AI and may contain mistakes.”
• “Hi! I’m an AI assistant for [Company]. How can I help?”

For an AI voice agent, at the start of the call:

• “Hi, this is [Company]’s automated AI assistant. This call may be recorded.”

Short, up front, unambiguous. The word “AI” or “automated” should appear in the notice itself — a robot avatar alone is a design hint, not information.

Wording that risks failing

• “Meet Sam, your assistant!” with a human-looking avatar — actively suggests a human. This is the opposite of disclosure.
• “Powered by AI”in a footer badge — ambiguous (many human-staffed tools are “AI-powered”) and often not “at the first interaction.”
• Disclosure only in the privacy policy or ToS— fails the “clear and distinguishable” and timing tests of Article 50(5).
• “Virtual assistant”alone — arguably implies software, but courts and regulators have not blessed it; adding “AI” costs you two characters.

Where to put it, by channel

• Web chat widget: pinned line in the chat header plus an AI-authored first message that self-identifies.
• WhatsApp / Messenger / SMS bots:the first outbound message identifies the sender as an AI assistant — you don’t control the surrounding UI, so the message itself must carry the disclosure.
• Voice: spoken disclosure in the greeting, before the agent starts handling the request.
• Hybrid human/AI support:disclose when the AI is talking and mark the handover (“Connecting you to a human agent…”). See is my chatbot covered? for whose duty this is.

Accessibility is part of the requirement

Article 50(5) ties the disclosure to applicable accessibility requirements. Concretely: the notice should be real text (not an image), have sufficient color contrast, and be announced to assistive technology — for a chat widget, ensure the header line and first message are in the accessibility tree.

Check the rest of your obligations

Disclosure wording is one of several Article 50 duties — synthetic content marking and deep-fake labelling have their own rules (see labelling AI-generated content). Run the free compliance checker to see which paragraphs apply to your product, and use the Article 50 checklist to get to compliant before enforcement (penalties reach €15M or 3% of worldwide turnover). Already have a GDPR privacy notice on the bot? It does not satisfy Article 50 — the two duties stack.