EU AI Act Article 50 explained for websites

Updated 18 November 2026

AI Disclosure: This content includes AI-generated text.

Short answer: Article 50 of the EU AI Act is the transparency rule, and it has applied since 2 August 2026. For a typical website it means three things: tell people when they are talking to an AI chatbot, visibly label realistic AI images, video and audio (deepfakes), and disclose AI-written text that informs the public on matters of public interest unless an editor reviewed it. The machine-readable marking of AI output is the job of the AI tool makers. This guide walks through Article 50 paragraph by paragraph and says what each one means for a website.

Checked on 18 November 2026 against the AI Act as amended by the Digital Omnibus on AI and the Commission’s Article 50 guidelines. Sources are at the end.

Article 50 at a glance

ParagraphWho it bindsWhat it requiresTypical website example
50(1)ProvidersAI systems that interact with people must tell them they are dealing with an AI, unless obviousChat widget, voice assistant, AI support agent
50(2)ProvidersAI-generated audio, image, video and text must be marked in a machine-readable, detectable wayImage generators, AI writing tools, AI features you offer to others
50(3)DeployersInform people exposed to emotion recognition or biometric categorisationRare on websites; webcam mood or age analysis
50(4)DeployersDisclose deepfakes, and AI text published to inform the public on matters of public interestRealistic AI photos, AI-written news or health articles
50(5)BothInformation must be clear, distinguishable, accessible, at the latest at first interaction or exposureWhere and how your notices appear
50(7)CommissionCodes of practice and guidelines on how to complyThe Code of Practice on AI-generated content

Provider or deployer: which are you?

Article 50 splits duties by role. A provider develops an AI system, or has one developed, and puts it on the market or into service under its own name. A deployer uses an AI system under its own authority in a professional context. Most website owners are deployers: they use ChatGPT, Midjourney or a chat widget made by someone else.

You can become a provider without building a model. If you put your own AI chatbot on your site under your brand, or sell a plugin or app with an AI feature, the Commission’s guidelines treat you as the provider of that system, even when it runs on another company’s model. Our article on responsibility for third-party AI tools goes deeper.

50(1): chatbots and other AI that talks to people

People must be told they are interacting with an AI system, unless it is obvious to a reasonably well-informed person from the context. Formally this is the provider’s duty, but in practice the website shows the notice, and many chatbot vendors leave it to you. A short line in the chat window before the first message is the usual solution, for example “You are chatting with an AI assistant. A human can take over on request.” Our chatbot disclosure guide has wording you can copy.

50(2): machine-readable marking

Providers of systems that generate synthetic audio, images, video or text must mark the output so that it can be detected as AI-generated. Systems already on the market before 2 August 2026 have until 2 December 2026 under the transitional rule added by the Digital Omnibus. As a deployer you have no marking duty, but it helps not to strip marks on purpose. Details are in our 2 December deadline guide.

50(3): emotion recognition and biometric categorisation

If your site analyses visitors’ faces, voices or other biometric data to infer emotions or sort them into categories, you must inform them and process the data in line with the GDPR. Very few ordinary websites do this. If yours does, get specific legal advice, because other parts of the AI Act may also apply.

50(4): deepfakes and AI-written text

This is the paragraph that affects most websites directly.

  • Deepfakes. If you publish AI-generated or manipulated images, audio or video that resemble real people, objects, places, entities or events and could falsely appear authentic, you must disclose it. For clearly artistic, creative, satirical or fictional work, a lighter disclosure that does not spoil the work is enough. See how to label AI images in WordPress.
  • Text. AI-generated or manipulated text published to inform the public on matters of public interest must be disclosed. The duty falls away if the text went through human review or editorial control and a person or organisation holds editorial responsibility. Product descriptions, marketing copy and fiction are generally outside this rule. See do you need to disclose AI-generated content.

50(5): how the information must be given

Every Article 50 notice must be clear and distinguishable, given at the latest at the first interaction or exposure, and must meet accessibility requirements. A paragraph in your privacy policy does not meet this test on its own. The notice belongs where the AI content or chatbot is. Our guide on where to place AI disclosures shows options for each case.

Need the actual words? The free AI disclosure generator writes a notice for chatbots, deepfake images, video or audio, and AI-written text, ready to copy as text or HTML.

Key dates

  • 2 August 2026: Article 50 applies, including chatbot notices and deployer labelling of deepfakes and public-interest text.
  • 2 December 2026: end of the grace period for machine-readable marking by generative AI systems that were already on the market before 2 August 2026.
  • 2 February 2027: the Code of Practice asks signatory providers to make watermark detection interoperable.

Penalties

Breaches of Article 50 fall under Article 99(4): fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher, and the lower of the two for SMEs and start-ups. National market surveillance authorities enforce it; in Germany that is the Bundesnetzagentur. More in EU AI Act fines and penalties explained.

Article 50 checklist for a website

  1. List every AI tool your site uses or shows to visitors: chatbots, AI search, image and text generators, translation, personalisation. Our AI usage audit guide helps.
  2. For each tool, decide whether you are provider or deployer.
  3. Add a notice to every chatbot, shown before the first message.
  4. Label realistic AI images, video and audio where they appear.
  5. Decide whether any AI-written text is public-interest information, and either disclose it or document the human editorial review.
  6. Check that notices are visible, readable and accessible on mobile.
  7. Write down who is responsible and review it when you add a new AI tool.

Our free EU AI Act Ready plugin covers steps 1, 3, 4, 5 and 7 on WordPress, with a readiness score and an exportable report.

Sources

This guide is general information, not legal advice. Written by the EU AI Act Ready team, 18 November 2026.