Over the past year many businesses have added an AI assistant to their website to answer customer questions, help choose products or book appointments. The AI Act does not forbid this, but it sets transparency rules: people must know when they are talking to a machine.
What does Article 50 require?
Article 50 contains four transparency obligations:
- Chatbots and voice assistants must be designed so that people are informed they are interacting with an AI system – unless this is already obvious.
- Providers of generative AI (e.g. image, video or text generators) must mark generated content in a machine-readable way so that it can be detected as AI-generated.
- Deployers of emotion recognition and biometric categorisation systems must inform the people exposed to them.
- Deepfakes and AI-generated text on matters of public interest must be disclosed, unless the text has undergone human editorial review and someone holds editorial responsibility.
What does this mean for a typical business website?
An AI chatbot or customer support assistant
This is the most common case. The simplest solution is to add a clear label to the chat window, such as “I am an AI assistant” or “Answered by AI”, and to say it in the first message as well. It is good practice to also offer a way to reach a human, for example by email or an enquiry form.
If you have a chatbot developed for you and use it under your own name, you may be the provider of the system under the regulation and responsible for meeting this requirement. If you use an off-the-shelf service, the service provider should ensure the disclosure, but it is your job to check that the label is visible and has not been removed.
AI-generated images and videos
If you use realistic AI-generated images or videos in your marketing that depict real people, places or events, they must be disclosed as AI-generated. For illustrative and clearly artistic images the requirements are lighter, but transparency is good practice anyway.
AI-assisted texts
Ordinary marketing copy, product descriptions and blog posts that have been reviewed and edited by a human generally do not need a separate label. The requirement mainly concerns texts on matters of public interest published without human editorial review.
What changed with the Digital Omnibus adopted in July?
The Digital Omnibus package adopted in July 2026 postponed the requirements for high-risk AI systems to December 2027, but the Article 50 deadline – 2 August 2026 – stayed the same. Generative AI systems already on the market before 2 August were given a transition period until 2 December 2026 to add machine-readable marking. Content created before 2 August does not need to be labelled retroactively.
What are the fines?
Breaching the transparency obligations can lead to fines of up to €15 million or 3% of a company’s worldwide annual turnover. Proportionality is taken into account for small businesses, but compliance is usually simple and inexpensive – so there is no reason to take the risk.
Checklist
- Does your website, online store or customer support use an AI chatbot or assistant?
- Can users see right at the start of the conversation that they are talking to AI?
- Is it possible to reach a human as well?
- Do you use realistic AI-generated images or videos, and are they labelled?
- Are AI-assisted texts reviewed by a person before publication?
- Does your privacy policy explain how chatbot data is processed?
Summary
Article 50 does not prevent you from using AI on your website – it requires honesty. A clear label in the chat window, a way to reach a human and well-considered privacy information are enough for most businesses. If you are planning an AI assistant or want to review an existing one, get in touch or see the AI solutions we build. You will also find practical use cases in our article AI solutions for small businesses.
Frequently asked questions
From when must an AI chatbot identify itself?
Article 50 has applied since 2 August 2026. The Digital Omnibus package did not postpone this deadline.
Do I need to label a blog post written with AI help?
Generally not, if the text has been edited by a person and someone is responsible for its content. The requirement mainly concerns texts on matters of public interest published without human editorial review.
What is the simplest way to comply?
Add a clear label in the chat window and the first message saying that the answers come from AI, and offer a way to reach a human. Describe the chatbot’s data processing in your privacy policy too.
We build websites, online stores and business software – and write about what we have actually built.
Learn more: AI solutions
Tell us what you want to achieve – we will reply within 3 business days with a concrete solution, price and timeline.
