The Quiet Deadline: What the AI Labeling Requirement Means for Content and Campaigns

While the industry debates the next AI model, something more practically relevant to content production and campaign work than any new chatbot is taking effect in just a few days: on August 2, 2026, the transparency obligations of Article 50 of the EU AI Act become effective. Hardly any other topic is being discussed as much right now – and hardly any as inconsistently.

What actually applies starting August 2

Article 50 requires that AI-generated or AI-manipulated content be made identifiable – concretely, this covers images, audio and video content, deepfakes, and chatbots, where users must be informed they're talking to a system. In practice, that means machine-readable labeling, for example via watermarks or signed metadata, and depending on the use case, a visible notice as well. This affects not just tech corporations but any company using AI-generated images in marketing, working with voice bots, or publishing AI-generated text on matters of public interest. The fines are no minor detail: up to 15 million euros or 3 percent of global annual turnover, whichever is higher.

The biggest misconception: "Now everything has to be labeled"

That's simply not true, and this misconception is currently causing more confusion than the law itself. Article 50 isn't a blanket labeling requirement for every AI use – it applies depending on context. Internal emails, drafts, or notes created with AI assistance aren't covered. Where a person reviews the content of a text and takes editorial responsibility by name, the labeling obligation falls away too. And a simple "Created with AI" label isn't sufficient in every case – but it isn't mandatory everywhere either. Whoever knows the law doesn't need to relabel every social media image; whoever doesn't know it ends up doing either too much or too little.

The point almost nobody in the CGI industry raises

This is where it gets concrete for our own craft: classic CGI – a product modeled from CAD data, digitally lit, and rendered, the way we've been doing it at Gieske Studios for years for catalogs and campaigns – isn't AI-generated image production in the sense of the article. It isn't created by a generative model but by classic 3D software and the decisions of an artist. The difference between CGI and AI image production is almost never made in the public debate, but it's exactly the point on which labeling obligations turn. As soon as generative AI tools enter the process – for variant creation or image completion, for instance, as we describe in AI-assisted content production – the classification becomes relevant and has to be checked case by case.

Why this isn't a reason to panic, but is a reason to act

The excitement around August 2 has two sides. One is justified: whoever hasn't documented at all which AI tools are in use in marketing, image production, or customer contact has catching up to do now – and there's a transition period until December 2, 2026 for systems already in use before the deadline, which should be used, not let slip by. The other side is overblown: the idea that from August every image, every text, and every social post has to carry an AI label creates exactly the kind of uncertainty a calm stocktaking would avoid.

The pragmatic first step looks the same for us – and for the companies we work with – every time: an inventory of which AI systems are actually in use, where they generate content for the public, and where a transparency obligation genuinely applies rather than a felt one. That stocktaking has meanwhile become part of our AI seminars for companies – not as a separate legal topic, but because it can't be separated from the question of which AI tools a company should sensibly use in the first place.

What to do now

Anyone who wants to act in the coming days shouldn't start with the technology, but with an overview: which AI applications are currently running in marketing, customer service, and content production? Which of them generate content that reaches the public? And where can labeling be solved technically in a simple way, for instance via metadata that many tools already provide anyway? That's a manageable task – as long as it isn't confused with the illusion that AI can no longer be used from August on. The opposite is true: the law demands transparency, not a ban.

Conclusion

August 2, 2026 is one of the least spectacular but practically most important deadlines currently occupying the creative and marketing industry. Whoever knows what Article 50 actually requires – and what it doesn't – doesn't need to panic or reinvent their own CGI production. Whoever doesn't know risks either a fine or unnecessary caution that hurts their own business more than the law itself.

This post is not legal advice and is not a comprehensive account of all obligations under the AI Act.