The setup check asks six short questions and scans your site for chatbots and AI plugins. Each answer is Yes, No or Not sure. You can change them at any time.
| Question | What it decides |
|---|---|
| Do people from the EU visit your site? | Whether Article 50 is likely to matter at all. |
| Is there a chatbot or AI assistant on the site? | The chatbot notice, Article 50(1). |
| Do you publish images, video or audio made with AI? | Labels for realistic AI media, Article 50(4). |
| Do you publish AI-written texts on news, health, finance, politics or other public topics? | Labels for public-interest text, Article 50(4). |
| Does a human editor review those texts and take responsibility for them? | Whether reviewed texts can skip the public label. |
| Do you use emotion recognition or biometric categorisation? | Article 50(3). Almost always “No” for a website. |
The result card
Each duty gets one of three marks:
- Applies: you should act on it. The card says what to turn on.
- Not yours: the duty belongs to the company that builds the AI (for example, machine-readable watermarks, Article 50(2)), or it doesn’t fit your answers.
- Your call: the law isn’t clear-cut for your case. A typical example is a chatbot whose vendor already shows a small notice.
The scan also lists AI tools it found, and widgets that are installed but not connected, so visitors don’t see them.
The result is a guide, not a legal opinion. If a mark surprises you, read the plain-English guides or ask a lawyer.