AI content labeling from August 2026: what do the new rules mean for e-shops?

07.08.2026
« Back to articles

Labelling AI-generated content from August 2026: what does it mean for e-shops?

From August 2026, new EU rules on labelling content created using artificial intelligence apply. What exactly do they mean for e-shops, and which AI-generated content actually needs to be labelled?

AI content labeling from August 2026: what do the new rules mean for e-shops?
AI content labeling from August 2026: what do the new rules mean for e-shops?

New rules for AI-generated content

Artificial intelligence is now a common part of creating texts, images and other content. From 2 August 2026 , however, new transparency rules under Article 50 of the AI Act began to apply in the European Union.

The rules focus primarily on situations where a person could mistake AI-generated content for something real or where AI informs the public about important societal issues.

What must be labelled under the AI Act?

The obligations differ depending on who provides the AI system and who uses it. Providers of generative AI systems must ensure that it is possible to identify generated or modified content using machine-readable labelling.

Companies and other professional AI users must, in certain cases, also inform the person directly who views the content. This applies in particular to deepfake images, videos and audio, as well as certain AI-generated texts that inform the public about matters of public interest. For interactive AI systems, such as chatbots, the provider must also inform users that they are communicating with AI, unless this is obvious from the circumstances.

Does an e-shop have to label every product description created using AI?

No. The AI Act does not introduce a general rule requiring every text created using AI to contain a label such as „Created using AI“.

The obligation to visibly label AI-generated text applies to texts published for the purpose of informing the public about matters of public interest. The European Commission includes, for example, politics, public safety, public health, environmental protection, or significant economic, scientific or cultural developments. Ordinary product descriptions, category descriptions or marketing texts therefore generally do not fall under this specific obligation, provided that their purpose is not to inform the public about a matter of public interest.

AI content labeling from August 2026: what do the new rules mean for e-shops?

What is good to know

Even text concerning a matter of public interest may not need to be labelled if it has undergone genuine human review or an editorial process and a specific person or company bears editorial responsibility for its publication. However, according to the Commission, merely checking spelling is not considered sufficient human review.

AI content labeling from August 2026: what do the new rules mean for e-shops?

AI images deserve greater attention

The situation is somewhat different with images. The AI Act uses the term deepfake not only for fake photographs of people. It can also refer to an AI-generated or modified image of an existing object, place, event or other entity if it may appear to a person to be authentic or truthful.

For e-shops, this may be important, for example, when creating realistic lifestyle photographs using AI. As one example, the European Commission mentions an authentic photograph of an empty apartment that was subsequently realistically furnished with furniture using AI. Such content may fall within the category of partially AI-modified content that needs to be transparently labelled. However, the specific context and whether the customer may regard the resulting image as an authentic depiction of reality are also always relevant.

What does this mean for e-shops?

If you use AI to create ordinary product descriptions, category descriptions or other e-shop content, the new legislation does not automatically mean that you need to add information to every text stating that it was created using AI.

Particular attention should be paid primarily to realistic AI-generated images and other content that a customer could regard as an authentic photograph of a real product, place or situation. For such content, it is necessary to assess whether it meets the definition of a deepfake and whether it is therefore necessary to inform the customer that AI was used. 

According to the European Commission, content generated and published before 2 August 2026 does not need to be labelled retroactively.

AI content labeling from August 2026: what do the new rules mean for e-shops?

What might the labelling look like?

The European Union has also prepared a set of icons for labelling AI-generated content. Their use is voluntary, and companies may choose another method as long as it meets the requirements of the AI Act.

If the labelling obligation applies to specific content, the information must be clear and recognisable to people no later than when the content is first displayed. For example, technical information hidden in a file alone is therefore not sufficient for deepfake content.

The new rules therefore do not mean the end of AI-generated content on e-shops, nor do they require everything that AI helped create to be labelled. What will be more important is knowing what type of content is created using AI, how it may affect the customer, and whether it falls among the cases to which transparency under the AI Act actually applies.

sources: heur-lex.europa.eu, digital-strategy.ec.europa.eu, asociace.ai

Pobo

Thinking about Pobo? Let us show it to you.

Book a short 30-minute call where we'll show you how Pobo works, what it's good for, and how it can save you a lot of time and money.

Book a call