An AI chatbot on your website — do visitors have to be told?
More and more websites run a chat widget where the answers come from an AI rather than a human. The EU AI Act puts a simple requirement on that situation: the visitor must be told that it is an AI they are talking to. The requirement sits in article 50.1 and applies from 2 August 2026 — but for most businesses it is as much a matter of trust as of law: a visitor who believes they are talking to a human and discovers otherwise rarely becomes a happier customer.
What does the law say?
Article 50.1 of the AI Act covers AI systems intended to interact directly with natural persons — chatbots are the textbook example. Such systems must be designed so that the persons are informed that they are interacting with an AI. There is one exception: when it is obvious to a reasonably well-informed person that it is an AI, it does not need to be stated explicitly.
The transparency requirements in article 50 apply from 2 August 2026. So it is not a requirement in force today — but it is a requirement with a date, and a chat solution bought or configured now will still be live when that date passes.
The more human-like the chat is — a name, a profile picture, casual language, typing indicators — the less obvious it is that it is an AI, and the clearer the disclosure needs to be. The exception for the obvious is meant for cases where no one could reasonably believe otherwise, not as a loophole for bots that work hard to appear human.
Who is responsible — the vendor or you?
The obligation in article 50.1 is formally directed at the provider of the AI system — whoever supplies the chat technology. But in practice the visitor's experience lands on you as the website owner: it is your website showing the chat, your brand in the header, and you who choose the tool and the configuration. Many chat platforms let you control whether the bot presents itself as an AI or appears under a human name.
The sensible position is therefore simple: make sure the chat on your website presents itself as an AI, regardless of where the formal provider responsibility sits. It is your customer interaction — and you are the one who gets the question if a visitor feels misled.
Which chats are affected in practice?
What matters is whether the visitor is having a conversation with an AI system. Some typical cases:
- A chat widget where the answers are AI-generated — should clearly show that the visitor is talking to an AI, for example "AI assistant" in the chat header or an introduction line such as "You are chatting with a digital assistant". Especially important when the tone is human-like.
- A plain FAQ search box without conversation — the visitor types a question and gets hits from a knowledge base — is normally not in the risk zone.
- A live chat staffed by humans is not an AI and is not covered — but if you combine a bot with human agents, it is the bot part that needs to present itself as an AI, and the handover to a human should be clear.
How CompliantHQ tests this
The scanner identifies chat and assistant widgets on the website — Intercom, Drift, Humany and others — and flags the transparency requirement in advance, so you have time to check with your vendor before the requirement becomes binding in August 2026. The check also runs during the trial period.
We cannot tell from a scan whether your particular chat is AI-driven or staffed by humans — only you and your vendor know that. The finding is therefore presented as something to verify, not as a confirmed violation.
How to fix it
- Find out how your chat works: ask the vendor whether the widget uses AI or automated answers, in whole or in part.
- If it is AI-driven — make sure that is clearly visible in the chat, for example "AI assistant" in the header or an opening introduction line. Most platforms have a setting for it.
- Don't give the bot a human name and a portrait photo without an AI disclosure — the more human the appearance, the clearer the disclosure must be.
- Check that the disclosure also works with screen readers, so it reaches every visitor.
- If the chat hands over to a human agent — make the transition clear, so the visitor knows when the AI stops and the human takes over.
What the check covers
- Chat and assistant widgets are identified (Intercom, Drift, Humany and more) and the transparency duty is flagged: visitors must be informed when they interact with an AI. Binding from August 2026.
Common questions
Do you have to say that it's an AI you're chatting with?
Yes — under article 50.1 of the AI Act, persons interacting with an AI system must be informed of it, unless it is obvious to a reasonably well-informed person. The requirement applies from 2 August 2026.
We use an off-the-shelf chat platform — isn't this the vendor's problem?
The obligation is formally directed at the provider of the AI system, but it is your website showing the chat and you who control the configuration. Make sure the bot presents itself as an AI — many platforms have a setting for it.
Is it enough that "chatbot" appears somewhere in the widget?
The point is that the visitor understands they are talking to an AI — before or at the latest when the conversation starts. A clear header such as "AI assistant" or an introduction line in the first message does the job; a tucked-away notice does not, especially if the bot otherwise behaves like a human.
What applies until August 2026?
The transparency requirements in article 50 apply from 2 August 2026. Nothing stops you from being transparent already — it costs nothing, builds trust and means there is nothing to change when the requirement takes effect. That is why we flag it in advance.
Want to see what we find on your site?
Run a free scan — all four modules included for 30 days, no card required.