EU AI Act: AI Content Labelling Rules Explained
AI-assisted and editorially reviewed · Responsible: Rainer Schmitt
AI Content and the EU AI Act: Transparency Rules Businesses Should Know
Since 2 August 2026, the transparency provisions of the EU AI Act have introduced new requirements for certain AI-generated and AI-modified content. The European Code of Practice on Transparency of AI-Generated Content is intended to help businesses translate these requirements into practical processes.
Agencies, marketing teams, online retailers and other organisations should therefore consider an important question: When must the use of artificial intelligence be disclosed, and which types of content are covered by the new rules?
Why does transparency for AI content matter?
Modern AI systems can generate texts, images, videos and audio recordings at a remarkably high level of quality. Users may no longer be able to determine whether a piece of content is authentic or has been created or altered with artificial intelligence.
The EU AI Act aims to improve transparency in this area. People should be able to recognise when they are interacting with an AI system or viewing content that has been artificially produced or manipulated.
Further information is available from the European Commission: Code of Practice for AI-generated content .
What is the purpose of the Code of Practice?
The Code of Practice on Transparency of AI-Generated Content was developed within the framework of the European AI Office. It contains practical recommendations for labelling content and making AI-generated material technically detectable.
One distinction is particularly important:
- Participation in the Code of Practice is voluntary.
- The legal provisions of the EU AI Act are binding.
- The Code can help organisations document how they are meeting their obligations.
Which types of AI content require particular attention?
Not every piece of content created with the assistance of artificial intelligence automatically requires a visible label. The rules focus on several specific use cases.
Deepfakes and manipulated media
Realistic images, videos or audio recordings that artificially depict or substantially alter people, places or events must generally be identifiable as AI-generated or AI-manipulated.
Automatically published texts on matters of public interest
Additional requirements may apply to AI-generated texts concerning matters of public interest. Examples include political, social, health-related or security-related topics.
Labelling may be necessary when such content is published automatically without human editorial oversight.
AI-powered conversational systems
Users should generally be informed when they are communicating with a chatbot or another artificial intelligence system. An additional notice may not be required where the nature of the interaction is already obvious.
Do AI-assisted blog articles have to be labelled?
There is no general obligation to label every text produced with the help of artificial intelligence.
A business may use AI to prepare an initial draft, structure an article or improve its wording. This does not automatically mean that the final publication requires an AI label.
What matters is whether a responsible person checks the information, edits the text and assumes editorial responsibility for publication. Content that is published fully automatically without meaningful review is more likely to raise transparency concerns.
How can businesses implement the requirements?
The use of generative AI should be governed not only through technical measures but also through internal policies and clearly assigned responsibilities.
Useful questions include:
- Which departments and tasks involve generative AI?
- Which content is reviewed by a person before publication?
- Who decides whether visible labelling is necessary?
- How is the use of AI documented internally?
- Are metadata, watermarks or other technical markers being used?
These processes can support regulatory compliance while also reducing the risk of publishing inaccurate, misleading or insufficiently reviewed material.
Technical labelling with metadata and watermarks
Technical markers can complement visible notices. These may include machine-readable metadata, digital provenance information and watermarks.
Such measures are intended to help platforms, search engines and other systems identify content that has been generated or modified with artificial intelligence. Businesses should therefore examine which labelling features are already supported by their AI tools and content management systems.
What does the EU AI Act mean for content marketing and SEO?
For content marketing, documented editorial responsibility is becoming increasingly important.
Artificial intelligence should not be treated as a replacement for professional expertise, quality control or a distinctive editorial perspective. It can accelerate workflows, support research and improve wording, but responsibility for the final content remains with the publishing organisation.
Strong SEO content should therefore continue to provide:
- accurate and verifiable information,
- clear value for the intended audience,
- an accessible and individual presentation,
- human review and editorial accountability.
Conclusion: Clear processes build trust and reduce risk
The transparency provisions of the EU AI Act are designed to make artificially generated content easier to identify and assess. The voluntary Code of Practice provides a practical framework that businesses can use when developing their own internal procedures.
The objective is not to label every use of AI without distinction. The relevant factors include the type of content, the way it is used and the extent of human oversight.
Companies that introduce reliable review procedures, assign clear responsibilities and document their use of generative AI will be better positioned to use the technology responsibly and transparently.