From 2 August 2026, content created with artificial intelligence has to be labelled across the EU. That includes advertising. It sounds more complicated than it is. Here is a plain overview of what the law wants and how it is handled in practice.
Key points
- From 2 August 2026 the EU AI Act requires synthetic content, including advertising, to be labelled.
- The label is machine-readable in the output and, for realistic content, visible to the viewer as well.
- Older generative systems have a transition period until December 2026. The code of practice was published in June 2026.
Quick answer
AI-generated video, photos and audio must carry a machine-readable marking that they were made with AI. For realistic “deepfake” footage (real-looking people, places, events) a visible notice for the viewer is added. The obligation applies from 2 August 2026.
What exactly the EU AI Act requires
The key part is Article 50 of the AI Act. Put simply, it says two things:
- A machine-readable trace. The provider of the AI system must make sure its outputs carry machine-readable information that they are artificial, typically in metadata; whoever publishes realistic AI footage is responsible for the visible label (the C2PA / Content Credentials standard).
- A visible label for deepfakes. If the content realistically depicts people, places or events that did not happen that way, the viewer must be able to tell from a clear and understandable notice.
Who it applies to
Not only big tech companies. The machine-readable mark is on the provider of the AI system. The visible label for realistic footage falls on everyone who publishes it: brands, online shops, agencies and production studios. It does not depend on whether the system is “high-risk”. If you put an AI ad out into the world, it applies to you.
From when, and in what steps
- 2 August 2026: the transparency obligations under Article 50 take effect.
- December 2026: generative systems placed on the market before 2 August 2026 have until then to add machine-readable marking.
- In June 2026 the European code of practice on marking and labelling was published. It is voluntary and specifies how exactly to label.
Dates according to the European Commission's quick facts on transparency rules for AI systems.
There are exceptions
The law takes context into account. For evidently artistic and creative works, the label can be shown in a proportionate way that does not get in the way of the work itself. But an ad that presents an AI shot as reality does not fall under the exception. There, you label.
How we handle it for you
You do not have to keep an eye on this yourself. For every output we:
- embed machine-readable metadata (Content Credentials) so the origin is traceable,
- add a visible label to realistic footage, in a format that suits the platform,
- and advise what to state where, so the campaign complies and still looks good.
Labelling is part of every job, whether it is an ad spot, a video with an AI avatar or product visuals.
Note
We are not a law firm, so take this text as an orientation, not legal advice. For bigger campaigns we recommend confirming the final labelling setup with your legal team. We are happy to supply the technical details.
Read next
- AI ads by well-known brands: how brands label AI content in practice.
- How an AI ad spot is made: the process step by step.
- AI video production and ad spots: EU AI Act labelling handled for you.
We will handle the labelling for you.
Planning an AI campaign and do not want to deal with compliance yourself? Get in touch. We will set it up so that it complies and looks right.
Book a call →