The website chatbot is the single most common AI deployment, and it is squarely inside Article 50’s first transparency case. If users can talk to an AI on your site, you owe them a disclosure — and it is one of the easiest obligations to satisfy.
What Article 50(1) requires
When a person interacts with an AI system they must be informed of that fact, in a clear and timely way, unless it is obvious to a reasonably well-informed person. For a chatbot that can be mistaken for a human agent, it is not obvious — so the disclosure is required. The bar is honesty at first contact, not a legal essay.
What good disclosure looks like
A short, visible line at the start of the conversation: “You’re chatting with an AI assistant. Ask to speak to a person any time.” Make it perceivable to assistive technology too — the AI Act expects disclosures to reach everyone, which aligns with WCAG accessibility. Avoid burying it in a privacy policy; the disclosure should be where the interaction happens.
The deadline and the proof
This obligation is enforceable from 2 August 2026. To be ready, confirm the disclosure is present and accessible, screenshot it, and date the record. If you swap chatbot vendors, repeat the quick check. That small, repeatable habit is what keeps you continuously compliant rather than scrambling before an audit.