Chatbot disclosure: UI examples that meet AI Act Article 50
Gideon Awolesi ·
Article 50 of the AI Act asks that people are told they're interacting with an AI system "in a clear and distinguishable manner at the latest at the time of the first interaction". Our AI disclosure templates cover what to say. This page covers where to put it in a chat interface, with five patterns you can combine, and what's unlikely to count.
01Pattern 1: a line above the conversation
Put a short line at the top of the chat window, above the first message, so it's the first thing people see when the chat opens. It works well for support chats that open in a panel, because the line stays in view while the conversation is short.
Keep it to one sentence. If it wraps onto three lines on a phone, people skip it.
02Pattern 2: the assistant introduces itself
Make the assistant's first message say what it is: "Hi, I'm [product]'s AI assistant. I can help with [tasks]." It's hard to miss, because it's in the conversation itself, where people are already reading.
Say what it can help with in the same message. It sets expectations, and it gives new users somewhere to start instead of a blank prompt box.
03Pattern 3: a label on every AI message
Label each message from the assistant, for example with "AI" next to its name. This matters in products where people and AI both reply in the same thread, like a support inbox where an agent can take over. Without a label on each message, people can't tell which replies came from a person.
04Pattern 4: a note next to the input
A short note under the input, like "AI can make mistakes, so check important answers", sits where people look before they type. On its own, it can be easy to miss, so pair it with pattern 1 or 2 rather than relying on it alone.
05Pattern 5: when a person takes over
If a person can take over the conversation, make the change visible. Show a line like "You're now chatting with Anna from our support team" and change the label on their messages. People treat a person's answer differently from an AI's answer, so they need to know which one they're getting.
06What's unlikely to count
- A notice in your terms of service or privacy policy only.
- A robot avatar on its own. The Commission's guidance reads the "it's obvious" exception narrowly.
- A tooltip people only see if they hover over something.
- A disclosure that only appears after the first answer.
07Make it accessible
The disclosure has to meet accessibility requirements too. Make sure screen readers announce it when the chat opens, give any AI icon a text label, use plain language with no abbreviations other than "AI", and don't make it a notice that disappears before people can read it.
If you'd like a second pair of eyes on your chat, send us one AI flow and one question for our free AI audit, and within five business days you'll get a recorded walkthrough and a one-page fix list.