AI-generated content and the EU AI Act: when do you need to label it?

Should a business disclose that content was created using AI?

Since 2 August 2026, Article 50 of the EU AI Act has applied transparency obligations to specific uses of artificial intelligence. The rule is more nuanced than saying everything created with ChatGPT needs an AI label.

Different obligations apply depending on whether an organisation is a provider or deployer, how the AI system is used and what type of content or interaction people are exposed to.

Why does the AI Act require transparency?

The basic principle is that people should be able to recognise certain situations where artificial intelligence is involved. Article 50 covers direct interactions with AI systems, machine-readable marking of certain AI-generated or manipulated content, emotion recognition and biometric categorisation, deepfakes, and certain AI-generated or manipulated text published to inform the public on matters of public interest.

The Commission describes these rules as measures intended to reduce risks such as misinformation, manipulation, fraud, impersonation and consumer deception, while helping people make informed decisions about AI interactions and content.

Chatbots should identify themselves

Providers of AI systems intended to interact directly with people must design and develop those systems so the people concerned are informed that they are interacting with AI, unless this is obvious to a reasonably well-informed, observant and circumspect person in the circumstances.

For a customer-facing AI assistant, the practical question is whether users clearly understand that they are communicating with AI rather than a human. Good transparency can be simple, for example: 'You are chatting with our AI assistant.' The objective is clarity rather than a lengthy legal notice.

What about AI-generated images, audio and video?

Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video or text content have machine-readable marking and detectability obligations under Article 50(2), subject to its scope and exceptions.

Deployers also have disclosure obligations for deepfakes. Under the AI Act, this concerns AI-generated or manipulated image, audio or video content resembling existing people, objects, places, entities or events in a way that would falsely appear authentic or truthful.

Businesses using synthetic spokesperson videos, cloned voices or realistic AI-generated imagery should therefore assess both their role and the nature of the content before deciding what transparency measure is required.

Does AI-generated text need a label?

Article 50 does not impose a visible label on every piece of AI-generated text. A specific deployer disclosure obligation applies where AI-generated or AI-manipulated text is published for the purpose of informing the public on matters of public interest.

That obligation does not apply where the content has undergone a process of human review or editorial control and a natural or legal person holds editorial responsibility for its publication. Other statutory exceptions can also apply in specific law-enforcement contexts.

Context matters. A short AI-assisted product description is not automatically equivalent to an AI-generated publication about a matter of public interest.

What counts as human review?

Businesses should be careful about interpreting human review too loosely. The Commission's current Article 50 questions and answers state that superficial, solely formal or procedural checks, such as spelling or grammatical correction, are not considered human review or editorial control.

A meaningful editorial process may involve reviewing material factual claims, checking important sources, correcting inaccuracies, assessing whether information is misleading and taking editorial responsibility for the final publication.

Simply selecting one AI-generated draft, clicking publish or correcting a spelling error should not be presented as substantive human review. Organisations should also avoid describing content as 'fact checked' or 'human reviewed' unless that process genuinely takes place.

Can businesses label more AI content voluntarily?

Yes. Article 50 establishes legal transparency requirements for specific circumstances, but an organisation can choose to disclose AI involvement more broadly where that supports trust and remains clear and accurate.

For example, an organisation publishing a substantially AI-generated Insights section can choose to identify every article consistently with a subtle 'AI-generated' disclosure. Blanche follows that approach for its Insights articles rather than attempting to create fragile distinctions between different levels of AI involvement.

What should an AI content disclosure look like?

Article 50 requires the relevant information to be provided in a clear and distinguishable manner, at the latest at the time of the first interaction or exposure, and to meet applicable accessibility requirements.

In practice, a visible disclosure should be understandable, positioned with the relevant content and available when a person first encounters it. Hiding the information only in a privacy policy or footer is unlikely to achieve the intended transparency.

The Commission has published a voluntary Code of Practice supporting Article 50 marking and labelling obligations. It has also released optional EU icons for labelling AI-generated content. Using the icons is not mandatory and does not establish legal compliance by itself.

Do AI providers and AI users have the same obligations?

No. Article 50 distinguishes between providers and deployers. Providers can have design and technical marking obligations, while deployers have disclosure obligations in particular situations including deepfakes and relevant text publications on matters of public interest.

Understanding the organisation's role for each AI system is therefore essential before deciding what transparency measures are required.

What should businesses review?

If your organisation creates, manipulates or publishes content using AI, a focused transparency review can identify where Article 50 or a voluntary disclosure approach is relevant.

• Which AI systems generate or manipulate content?
• Is the organisation a provider, deployer or both for the relevant system?
• What types of content are produced and who sees them?
• Are realistic synthetic images, audio or video involved?
• Does AI-generated or manipulated text inform the public on matters of public interest?
• What substantive human review or editorial control takes place, and who holds editorial responsibility?
• How are required disclosures displayed and documented?

Transparency is part of good AI governance

AI disclosure should not be treated as an isolated website issue. It connects with AI policies, content verification, human oversight, accountability and the organisation's wider inventory of AI use.

Organisations increasingly need to know not only where they use AI, but also when customers, employees or the public should be told about that use. Article 50 makes that a practical governance question.