←All insights

Guide

The AI Act and Your Chatbot: What Do You Need to Include on Your Website?

Starting August 2, 2026, the European AI Act requires every website with a chatbot to clearly inform visitors that they are talking to an AI and not a human. This requirement is set forth in Article 50 and also applies to self-employed individuals.

July 14, 2026

A smartphone displays a chat conversation with an AI assistant, along with a banner noting the mandatory AI disclosure requirement effective August 2, 2026, under the AI Act

You have a chatbot on your website. It answers questions, helps visitors get started, or collects information for a request.

That's handy. But does your visitor know they're talking to an AI?

As of August 2, 2026, the transparency rules set forth in Article 50 of the European AI Act have been in effect. For AI chatbots, this boils down to one clear message: visitors must be able to recognize that they are communicating with an AI system.

Below, I'll explain what to check, which notification to use, and why not every image or text you create with AI follows the same rules.

What Does the AI Act Mean for Your Chatbot?

Article 50 requires the provider to disclose information regarding AI systems that communicate directly with people. An exception applies when the AI nature is already clear within the context of use.

Do you use an external chatbot tool? If so, ask the provider how it complies with that disclosure requirement.

Are you having a custom chatbot built, or are you offering a system yourself? If so, you’ll also need to determine exactly what your role is. That depends on how the system is developed and offered.

For your website, the practical test is simple: open the chat just as a new visitor would. Is it immediately clear that an AI is responding here?

What message will you include with your chatbot?

I would just call it out:

“Hello! I’m the AI assistant for [company name]. How can I help you?”

That’s clearer than just “virtual assistant.” That description doesn’t directly indicate that AI is behind it.

Next, check how the notification appears. The information must be clear, distinguishable, and accessible, no later than the first interaction.

So be sure to test it on your cell phone as well. Does the opening message disappear behind a button? Does the user have to scroll first? Does the notification remain visible when the call starts?

A name, avatar, or friendly writing style can be nice. Just make sure the visitor always understands who—or what—is responding.

What should you ask your chatbot provider?

You don't need to know the entire law by heart to do that. Start with these questions:

  • How does the chatbot make it clear that it is an AI system?
  • Can I edit the introductory message myself?
  • How does the notification work on mobile devices and with assistive technology?
  • Who checks this when the tool is updated?
  • What rules apply to the information that visitors enter?

Save the answer in your website's documentation. That way, you'll know later which settings and agreements you've checked.

Do you have to label all AI-generated images?

Not automatically. Article 50 distinguishes between technical labeling by providers and visible information accompanying certain publications.

For users, the reporting requirement for deepfakes is particularly relevant. This requirement does not apply only to people: images, audio, or video of existing places, objects, or events may also be subject to it if they appear to be real when they are not.

So don't just go by the question: “Is there a face in it?”

Also consider what your image conveys. For example, does a generated image look like a real photo of your business, product, or completed project? If so, it deserves extra attention.

Regardless of the legal assessment, I think it's important that your visitors don't get the wrong impression of what you're actually offering.

What about texts written using AI?

There is no general requirement to include the phrase “written with AI” at the end of every text.

The specific reporting requirement applies to AI-generated or AI-edited texts that are published to inform the public about topics of general interest. An exception applies in cases of human review or editorial oversight, provided that someone also bears editorial responsibility.

My approach, therefore, remains the same: check what you publish. Are the facts accurate? Do the examples apply to your business? Do they contain promises you can’t keep?

That's necessary even without a labeling requirement.

Can you get a fine?

Article 50 is subject to the penalty provisions of the AI Act. The general maximum is €15 million or 3% of global annual revenue; for SMEs, the lower of these two thresholds applies.

That is a maximum amount, not a standard fine. The specific assessment depends, among other things, on the circumstances of the violation.

That’s why I wouldn’t make changes to your website out of panic. Rather, you should do it because visitors deserve clear information, and you want to know how your chatbot is set up.

Checklist for Your Website

  1. Open your chatbot as a new visitor.
  2. Check to see if the AI notification is immediately clear.
  3. Test the display on desktop and mobile.
  4. Ask your supplier how the transparency rules are implemented.
  5. Check separately which rules apply to your AI images, audio, video, and text.
  6. Make a note of what you've checked and which questions are still open.

Need help?

Not sure where to edit the welcome message? Or do you want to check how your chatbot appears on your website?

In that case, I’d be happy to review the settings and technical implementation with you. For a legal assessment of your specific application, I recommend that you consult a specialized attorney.

Let me know which chatbot you use →

Would you like to read more about AI-generated content and the transparency rules? Sirius Legal Advocaten explains what Article 50 means for deepfakes and marketing.


Frequently Asked Questions

Does my chatbot have to disclose that it's AI?

With an AI chatbot, visitors need to know they’re communicating with AI, unless that’s already clear from the context. This information must be clear and accessible no later than the first interaction. For example, use: “Hello, I’m the AI assistant for [company name].” Are you using a third-party tool? Ask your provider how they handle this and check the disclosure on your website.

Does the AI Act also apply to small businesses and the self-employed?

Yes, being a small business does not grant a general exemption from Article 50. The obligations that apply to you depend on your specific use case and your role. Regarding the disclosure requirements for AI chatbots, Article 50 focuses on the provider. If you use an external chatbot, you should therefore ask the provider how it complies with these requirements.

Do I need to label all my texts when I use AI to write?

No. Article 50 does not impose a general labeling requirement on every text you create using AI. The specific rule applies to AI-generated or AI-edited texts that you publish to inform the public about matters of general interest. An exception applies in cases of human oversight or editorial supervision, when a person or organization bears editorial responsibility. Therefore, simply putting your name under a text does not fully satisfy the requirement.

Guide

July 14, 2026

Next insight

Would you rather do it than read about it?

Tell me where it is
freezes.

You don't need to know which tool you need. Just tell me where you're having trouble. Then we'll figure out together how to do it more efficiently.

Let's build→