If your Intercom, Zendesk, Drift or Crisp widget answers visitors with AI, the people using it need to be told they are dealing with an AI, at the latest at their first interaction. Don't assume the vendor has switched that on for you. Intercom's own documentation says its "AI Agent" label in the Messenger is off by default for customers who set up Fin after 10 July 2025. The fix takes a few minutes per tool: give the bot a name that doesn't pass as human, add a disclosure to its first message, and turn on any built-in AI label.
Key point: Article 50(1) of the EU AI Act requires that people are informed they are interacting with an AI system, unless that is already obvious. Article 50(5) requires the information to be clear and distinguishable and given at the latest at the first interaction. A vendor chat widget with a human first name and no AI label does not meet that on its own.
Why this is your problem, not only the vendor's
Formally, Article 50(1) is a provider duty, and for an off-the-shelf AI agent the vendor is most likely the provider. In practice, you control what visitors see: the bot's name, avatar, greeting and which settings are switched on. If the widget runs on your site as "Sarah from Support" with no mention of AI, the undisclosed AI is on your website, under your brand. For the full provider and deployer analysis, see do I need Article 50 compliance if I only use a chatbot?
What a good disclosure looks like
Whatever the tool, aim for three things:
- A name that doesn't pass as human. "Fin AI Agent" or "Support Assistant (AI)" rather than "Sarah". Law firms report that the Commission's July 2026 guidelines reject generic references to an "assistant" as a disclosure on their own, so say "AI" explicitly.
- A disclosure in the first message, before the visitor has to ask. For example: "Hi, I'm an AI assistant. I can answer most questions, or connect you with a person."
- A disclosure that survives the handover. If the conversation moves between the AI and a human agent in the same window, make it clear who is answering at each point.
Don't rely on a line in your terms of service or privacy policy. For wording you can adapt, use the free AI chatbot disclaimer template. For placement options, see Article 50(1) chatbot disclosure, explained.
Intercom (Fin AI Agent)
Intercom has built-in disclosure settings. Check them, because two of them depend on when you set up Fin.
- "AI Agent" label in the Messenger. In Intercom, go to Fin AI Agent > Fin settings > Identity and turn on Show AI Agent label in Messenger. When it's on, an "AI Agent" label appears after Fin's name on its replies. Intercom's documentation says this is on by default for workspaces that deployed Fin before 10 July 2025, and off by default for newer ones.
- Introductory message. Intercom offers an optional intro message that says the visitor is speaking with an AI Agent and can ask for the team at any time. Turn it on, or write your own.
- Email footer. Under Fin AI Agent > Fin settings > General, "Show 'Composed by AI' in email footer" adds a line to Fin's email answers. Intercom says this is on by default. Leave it on.
- Fin's name. If you renamed Fin to a human first name, change it back or add "AI" to it.
Zendesk (AI agents)
Zendesk's documentation doesn't describe a built-in AI label shown to customers in the messaging widget, so the disclosure has to come from the agent's name and its first message.
- Name and avatar. In Admin Center, open your AI agent's settings and find its Identity section. The name appears at the top of the Web Widget in messaging conversations and as the sender in emails. Zendesk's own advice is to "choose a name that makes it clear customers are not talking to a human".
- First message. Add a clear AI disclosure to the greeting or first reply your AI agent sends.
- Fin for Zendesk. If you run Intercom's Fin inside Zendesk, Intercom says its AI Agent disclosure can be configured there too.
Drift (Salesloft)
Drift is now part of Salesloft and is sold with an AI chat agent. Its chat is built from bots with a name, avatar and welcome message, so the same approach applies.
- Bot name and avatar. Make sure any bot that sends AI-generated answers has a name that says it's an AI, not a human persona.
- Welcome message. Put the disclosure in the first message the bot sends on each page it runs on, including playbooks aimed at specific pages or campaigns.
- Handover. Drift often routes qualified visitors to a sales rep in the same window. Make it clear when a person takes over.
Crisp (Hugo AI agent)
Crisp's AI agent is called Hugo. Crisp's documentation lets you set the name and avatar visitors see, but doesn't describe an automatic AI label, so again the disclosure has to be added by you.
- Name and avatar. Under Agent → Settings in Hugo, set a name that makes clear it's an AI. If you keep the name "Hugo", add "AI" to it or say so in the first message, because "Hugo" on its own reads as a person.
- First message. Add the disclosure to Hugo's greeting or first reply.
- Handover. Crisp can route conversations between Hugo and your team in the same inbox. Make it clear to the visitor when a human replies.
Check it the way a visitor sees it
Settings pages don't tell you what visitors actually get. After making changes:
- Open your site in a private browser window, open the chat and send a question. Is there a clear AI disclosure before or with the first answer?
- Check the bot's name and avatar in the conversation itself, not just in the admin panel.
- Trigger a handover to a human, and check that the switch is visible.
- Check your email channel too, if the same AI answers emails.
- Repeat on mobile, where some widgets show a shorter layout.
- Check again after vendor updates or plan changes, since new features can arrive with new defaults.
What the Article50.io scan can and can't see
A free Article50.io scan loads one public page, detects Intercom, Drift, Zendesk or Crisp by their scripts (or custom chat markup), and looks for AI-disclosure wording on the page and inside the chat widget as it first loads. It doesn't open the chat or start a conversation, so a disclosure that only appears once a conversation starts, such as Intercom's AI Agent label on Fin's replies, can't be seen, and the scan may still flag the page. Treat a flag as "no disclosure is visible before the chat is used", then open the chat and check it yourself. A clean scan does not mean you are compliant.
Frequently asked questions
Is Intercom's AI Agent label switched on by default?
It depends on when you set up Fin. Intercom's documentation says the "Show AI Agent label in Messenger" setting is on by default for workspaces that deployed Fin before 10 July 2025 and off by default for newer ones. You'll find it under Fin AI Agent > Fin settings > Identity.
If my chatbot vendor is the provider, do I still need to do anything?
Yes, in practice. Article 50(1) formally places the duty on the provider, which for an off-the-shelf AI agent is most likely the vendor. But you choose the bot's name, avatar, greeting and settings. If visitors can't see that they are talking to an AI, the gap is on your website and under your brand.
Is a bot name like "Hugo" or "Sarah" enough of a disclosure?
No. A human-sounding name suggests a person, not an AI. The exception in Article 50(1) only applies where it is obvious to a reasonably well-informed, observant and circumspect person that they are dealing with an AI. Add "AI" to the name, say so in the first message, or both.
Where should the AI disclosure appear in a chat widget?
Where the visitor sees it at the latest at their first interaction: in the bot's name, its first message, or a label on its replies. Article 50(5) requires the information to be clear and distinguishable. A line in your terms of service or privacy policy is unlikely to be enough.
Does a rules-based chatbot need a disclosure?
Article 50(1) covers AI systems. A purely scripted decision-tree bot may fall outside it. Many vendor widgets now include AI answer features, though, and a new plan or feature can switch them on. Check whether your widget generates answers with AI before deciding you don't need a disclosure.
This article is general information, not legal advice.