AI-generated content: when to label it?
Do you use AI for product descriptions, blogs or images? A label is not always required. The AI Act distinguishes between makers of AI tools and users such as web shops.
Article 50(2) applies to providers of generative AI, such as the makers of ChatGPT or Midjourney. They must mark the output in a machine-readable way, for example with C2PA Content Credentials, IPTC metadata or a watermark. A web shop that uses such a tool is usually not a provider.
As a user (deployer), Article 50(4) applies. You must label deepfakes: AI images, video or audio that realistically resemble real persons, products, places or events and appear authentic. For evidently artistic or satirical content, a disclosure that does not hamper the work is enough.
You must label AI text that informs the public on matters of public interest, such as news, health, finance or legal information, unless it has undergone human editorial review. Ordinary product descriptions normally do not fall under this.
Practical tip: even where it is not required, a label builds trust. Preferably use the official EU icons from the June 2026 Code of Practice, for example 'Made with AI' next to an AI image.
Want to know if this applies to your website?
Our scanner automatically checks whether your website meets these AI Act requirements and gives concrete recommendations.
Scan your website