AI labelling on websites: What Article 50 means for editorial teams
The transparency requirements set out in Article 50 of the EU AI Regulation have been in force since 2 August 2026. Anyone currently looking into this topic might quickly get the impression that as soon as AI is involved in any context, a label must be attached.
It’s not quite that straightforward.
For website operators and editorial teams, it depends on how AI is used, what content is generated as a result, and whether this content undergoes editorial review prior to publication. So what really needs to be labelled and what doesn’t? We take a look at the key scenarios for websites.
AI-generated texts: Does a disclosure now have to be included everywhere?
No. When it comes to texts, labelling is particularly relevant when AI generates or modifies content on topics of public interest and this is published without human review. If, on the other hand, an AI-generated text is fact-checked and editorially vetted, an exception to the disclosure requirement applies.
For editorial teams, the key question is therefore not: ‘Did AI help write this?’, but rather: ‘Who checked the content and is responsible for it?’
AI images: What needs to be labelled?
The same applies to images: simply being created by AI does not automatically mean they must be labelled. So-called ‘deepfakes’ are particularly relevant. These are AI-generated or manipulated images that resemble real or plausibly realistic people, places, objects or events and could be mistaken for the genuine article.
Labelling:
- photorealistic AI characters that appear to be real people
- manipulated photos depicting a situation that did not actually take place
- realistic-looking AI images of places or events
Generally do not require labelling:
- obviously artificial illustrations, icons or graphics that cannot be mistaken for an authentic depiction
- technical image optimisations such as sharpening or noise reduction
- Edits that do not significantly alter the message or meaning of the original
A simple check for editors: Could someone mistake the result for an authentic depiction? If so, the labelling requirement should be examined more closely.
A relevant example: Our featured image is also AI-generated and deliberately exaggerated. Of course, nobody sticks ‘AI’ labels on their website. Nor does the legal principle apply that ‘if it contains AI, it must be labelled’. The image depicts a staged symbolic scene and is not intended to be a supposedly genuine documentary.
Chatbots: Make it clear who is replying
With AI chatbots, the matter is relatively straightforward: users must, as a matter of principle, be able to recognise that they are interacting with an AI system and not with a human being. If this is obvious anyway, additional information may not be necessary.
Automatic translations: A less obvious case
Automatic translations may also be affected if texts of public interest are automatically translated and published without editorial review. This is precisely what we have implemented at chemnitz.de. The English and Czech content there is automatically translated using DeepL via our TYPO3 ‘autotranslate’ extension.
Since August, a note indicating that the content has been machine-translated has been automatically displayed on these language versions. This labelling is integrated centrally within TYPO3. New pages, news items and events therefore receive this labelling without any additional effort on the part of the editorial team.
For us, this is the more sensible approach: if the content process is automated, the necessary transparency should also be automated.
What should editorial teams do now?
The transparency requirements have been in force since 2 August 2026. Rather than applying AI labels everywhere as a precautionary measure, it is therefore worth taking a quick look at your own processes:
- Where does AI generate or modify content?
- Which of these is subject to editorial review before publication?
- Where do users interact directly with AI?
- Where is labelling actually necessary?
If labelling is required, it should be displayed directly alongside the relevant content and not hidden away somewhere in the legal notice.
Not every use of AI requires a label. But every editorial team should know where and how they are using AI.
We help to categorise existing AI applications on websites and integrate the necessary transparency directly into CMS and content processes.
