AI Act marketing is no longer a topic for the future. Since 2 August 2026, Article 50 of the European AI Regulation sets transparency duties for certain AI content. For marketing teams that create images, videos or texts with AI, this raises a practical question: what has to be labelled and what does not?
As Marketing Director of a fashion brand, I lead a team of 22 people across several markets. AI is part of our toolbox, so the question is very concrete for me. This post is not legal advice but an orientation from practice.
- Article 50 of the AI Regulation has applied since 2 August 2026.
- Deepfakes must be disclosed, in advertising too.
- For artistic or fictional works, a disclosure that does not disturb the work is enough.
- Every campaign needs a clear team rule and, when in doubt, the legal team.
Contents
What Article 50 covers
Regulation (EU) 2024/1689 distinguishes between providers who develop AI systems and deployers who use them. Providers of generative AI must mark their output in a machine-readable way as artificially generated. Deployers, including brands and agencies, must disclose when they publish deepfakes (see here: EUR-Lex, Regulation (EU) 2024/1689).
For AI Act marketing, the deployer role matters most. The tool usually supplies the marking in the file, while disclosure to the audience remains the brand’s job.
Timing matters too. The regulation entered into force in 2024, but its duties apply step by step. For AI Act marketing, the key date is August 2026, when the transparency rules for content apply. Campaigns published after that should already meet the rules, even if the visuals were made earlier.
What counts as a deepfake
By deepfake, the regulation means image, audio or video content that resembles real people, objects, places or events and falsely appears authentic. An AI model that looks like a real person, or an invented scene at a real place, can fall under it.
A clearly illustrated visual or an obvious fantasy world, by contrast, is usually not a deepfake. This matters in fashion, because AI images here are often meant to look photographic. Where AI images work in fashion and where they do not is the topic of my post on AI image generation in fashion.
AI Act marketing for campaign visuals
For artistic, creative, satirical or fictional works, Article 50 provides some relief: disclosure should be made in a way that does not hamper the display of the work. A short, clearly visible note in the visual or in the caption can do this.
Whether an advertising visual counts as a creative work is a question of interpretation. The European Commission is working on guidelines and a code of practice. Until a practice has settled, I would rather label once too often than once too rarely.
For AI Act marketing day to day, I recommend a simple list per campaign: which visuals were made with AI, do they show real people or places, and where does the note go? The list takes a few minutes and gives clarity if someone asks later.
What applies to AI texts
For texts, the duty is narrower. Disclosure is required for AI texts published to inform the public on matters of public interest. It does not apply when a human has reviewed the text and holds editorial responsibility.
For most product copy and campaign lines this is unlikely to be the case. Still, a clear team rule pays off: every AI text is reviewed by a person, and the reviewer is named in the workflow. What this looks like in production is covered in my post on campaign production with AI.
Agencies should know these rules as well. Anyone supplying visuals or texts should state where AI was involved, so the brand can decide correctly as the deployer. AI Act marketing is therefore also a topic for briefings and contracts.
My conclusion on AI Act marketing: the rules are less strict than feared, but they require order. Deciding today when to label and who reviews lets you use AI with a clear conscience and protects trust in the brand.