A chatbot on your site: what must you tell the visitor?
Updated: 13 October 2026
The visitor should learn that they are talking to an AI program before the conversation starts. Since 2 August 2026 that duty has rested on the provider of the chatbot, not on every website that embeds one. Your part is usually to make sure the notice is not switched off or hidden.
Who has the duty
The EU Artificial Intelligence Act requires AI systems that talk to people to be designed so that a person knows they are dealing with AI. This provision has applied since 2 August 2026, with no transitional period. It is addressed to the provider of the chatbot, meaning the company that built it and offers it.
It covers chatbots and voice assistants based on AI. It does not cover simple auto-replies with fixed answers that are not AI, or a human agent who uses AI suggestions but sends the reply personally.
The detail below comes from the European Commission guidelines. The guidelines are not binding, and their formal adoption in all languages could not be confirmed by 13 October 2026. The voluntary Code of Practice on labelling content does not cover chatbots.
When the website itself is the provider
If your organisation built the chatbot itself and put it into service under its own name, it is the provider. The same applies if you take an existing system, modify it substantially, for example with new training data, and run it under your own brand.
One point is not decided: whether a site becomes a provider when it only configures someone else's chat widget, that is, sets the instructions, the look and the knowledge base. The guidelines do not say so expressly. The riskiest case is a bot running under the site's own name.
The "it is obvious" exception is narrow
No notice is needed when it is obvious to a reasonable, observant person that they are talking to AI. The Commission reads this exception narrowly: there must be almost no doubt left.
The guidelines expressly exclude chatbots embedded in online platforms and helpdesk tools from it, because users may take their replies as written by a human. So a chatbot on a public website has to give the notice.
What the notice should look like
One prominent notice before the first exchange is usually enough, for example "You are interacting with an AI system". It can be a line next to the chat box, the bot's first message or a permanent badge near the input field.
The bot should also answer truthfully when asked whether it is a human. In conversations about health, finance, legal matters or complaints it is worth repeating the notice.
These are not enough: a mention in the terms and conditions only, the bare word "assistant", a human name and photo, or a general line such as "services on this website use AI". The notice has to be clear, distinguishable and accessible to screen readers too.
What to do when you embed someone else's chat
Check that the supplier's widget shows an AI notice before or with the first message. Do not switch it off, hide it or restyle it until it disappears.
Do not give the bot a human name, a photo or an "agent" persona without a clear statement about AI. Make sure the notice is in the language of the page and right by the chat window.
A sentence about AI in the cookie consent window or the privacy policy does not replace the notice in the chat. At most it can complement it.
If the bot creates text or images for the visitor, the hidden marks in them are the tool maker's job. If you then publish that output yourself, it is assessed separately, under the rules for texts and images.
This describes the rules. It is not advice on your particular case. Legal basis: Article 50(1) and (5) of Regulation (EU) 2024/1689.
Key points
- The duty to tell people about AI lies with the chatbot provider, since 2 August 2026.
- A website becomes the provider when it built the bot itself, or substantially modified someone else's and offers it under its own brand.
- Whether merely configuring a third-party widget makes the site a provider has not been decided.
- The "it is obvious" exception does not cover helpdesk chatbots on websites.
- Do not switch off the AI notice in the widget. A line in the cookie consent window does not replace it.
Sources
- Regulation (EU) 2024/1689, the Artificial Intelligence Act
- European Commission: guidelines on Article 50, annex to document C(2026) 5054
- European Commission: page for the guidelines on transparency obligations for providers and deployers of AI systems
- European Commission: FAQ on transparency obligations under Article 50
This guide is for information only and is not legal advice.