Must your chatbot disclose it's AI? Yes, from 2026

It's official: from 2 August 2026, your chatbot must disclose that it's AI. Got a chatbot on your website, a WhatsApp bot, or an AI phone assistant handling customers? The European AI Act says you have to let people know they're talking to a computer, not a human. Sounds obvious, but almost no small or medium business does it yet, and most owners don't even know it's required. The good news: it's one of the easiest rules in the entire law to comply with. This article explains exactly what's required, what does not count as disclosure, and how to sort it in ten minutes.

What does the law actually say?

The rule is in Article 50 of the AI Act, the European law on artificial intelligence. The core is one sentence: when an AI system communicates directly with people, those people must be clearly told they're dealing with AI. It doesn't matter whether you built the chatbot yourself or bought it off the shelf, the moment it sits on your website or phone line, you're responsible for making sure the disclosure is there.

That obligation takes effect on 2 August 2026. From then on it's not 'good manners', it's law. In the Netherlands the Dutch Data Protection Authority (AP) supervises it. In May 2026 the European Commission published detailed guidance setting out exactly how strict it's meant to be, and it shows that a lot of companies are still doing it wrong.

Why does this rule exist?

The thinking behind it is simple: people have a right to know whether they're talking to a human or a machine. That used to be obvious, but now that chatbots sound more and more human, it isn't anymore. Someone who thinks there's an employee on the line might share information or make a decision they wouldn't otherwise. The disclosure rule protects that trust. And that's exactly why it works in your favour too: a customer who knows they're talking to AI has realistic expectations and is less easily disappointed when the bot doesn't know something.

Does this apply to my website chatbot or helpdesk bot?

Almost certainly yes. A lot of owners think: 'my bot only answers simple questions, surely it doesn't count.' But the law makes no exception for 'small' or 'simple' bots. A chatbot that handles FAQs, a bot that books appointments, an AI phone assistant that picks up your calls, they all fall under it. The European Commission even names helpdesk and customer service bots as an explicit example.

There's one exception: if it's blatantly obvious to a normal, attentive person that they're talking to AI, you don't have to disclose it separately. But don't bank on that too quickly. Modern chatbots sound more human all the time, which is precisely why the rule exists. 'It was obvious anyway' is a risky bet. The safe route is simple: just disclose it.

What does NOT count as a valid disclosure?

This is where a lot of companies go wrong. They think they've handled it, but their way of disclosing doesn't qualify. According to the European Commission's guidance, these do not count as a valid disclosure:

• A line buried in your terms and conditions or privacy statement. Nobody reads those before asking a question, so it doesn't count.
• Just giving your bot a name like 'Anna' or 'Dan'. A friendly name actually suggests it's a human. That backfires.
• The word 'assistant' or 'digital colleague' with nothing more. That doesn't say it's AI, an assistant can be a person too.
• A notice that only appears after the customer has already had a whole conversation. Too late: the customer must know before or at first contact.
• Small grey text that blends into the background. The disclosure has to stand out, not be hidden.

So how do you disclose it properly?

It really doesn't have to be complicated. A good disclosure is clear, timely and visible. A few ways that do qualify:

• An opening message in the chat: 'Hi! You're talking to our AI assistant. Question too tricky? I'll hand you over to a colleague.'
• For an AI phone assistant, a short line at the start of the call: 'You're speaking with the digital assistant of [company].'
• A visible label in or near the chat window, for example 'AI chat' or 'Automated assistant', that the customer sees before they start.
• The option to switch to a human. Not a legal requirement for the disclosure itself, but it makes the experience honest and prevents frustration.

And AI-generated text, images and video?

Article 50 covers more than chatbots. If you create AI-generated content, text, images, audio or video you put out, in many cases it must also be recognisable as artificially produced. For so-called deepfakes (realistic-looking AI images of people or events) it must be clearly labelled. For average SME marketing this is less strict than the chatbot rule, but it's good to know it's coming.

What happens if you don't?

The maximum fines in the AI Act are hefty: up to 15 million euros or 3% of your worldwide annual turnover for breaching the transparency rules. But don't panic. Those maximums are aimed at large players who deliberately ignore the rules. For an SME the fine is proportionate: with a turnover of, say, 2 million euros you're talking about a far lower amount, not millions.

The real risk for most owners isn't the mega-fine, it's the awkwardness: a customer who feels misled because they thought they were talking to a person, a complaint to the regulator, or bad press on social media. And that's a shame, because one sentence prevents it.

The easy route: build the disclosure in from the start

If you already have a chatbot or AI phone assistant, the disclosure is often added in a few minutes, a line of text in the opening message or a label by the chat window. If you're having a new solution built, make sure the disclosure is in there from day one. Then 'complying with the AI Act' isn't a separate project, it's just part of how the thing works.

At Socialo I build customer-communication automation so the AI disclosure is in there by default, clear, at the right moment, and with a smooth handover to a human. Not because I have to, but because being honest about what's AI simply works better for your customer relationship.

What should you do now?

• List every place where AI talks directly to your customers: website chat, WhatsApp, phone, social media DMs.
• Check each one: is there a clear disclosure that it's AI, before or at first contact?
• Don't hide the disclosure in your terms or in grey mini-text, put it in the conversation itself.
• Make sure customers can switch to a human if they want to.
• No chatbot yet but want one? Build the disclosure into the design straight away, so you're compliant from the start.