AI Already Makes 90% of Zalando's Ads — Law Lags Behind

Eurocommerce, with members Amazon, H&M and Ikea, asks to exempt AI product ads from the EU AI Act's labeling rule before it takes effect on August 2.

4 min readEAEvgenii ArsentevEvgenii Arsentev · PhD

Europe's largest retail trade association is asking Brussels to exempt AI-generated advertising from the EU AI Act's labeling requirement, weeks before that rule takes effect. Eurocommerce — whose members include Amazon, H&M, Inditex and Ikea — argues that an AI-made image of a product, like a sofa staged in a living room, shouldn't count as a 'deepfake' and therefore shouldn't need an AI-generated label. The group's director general, Christel Delberghe, made the case in a letter to EU tech commissioner Henna Virkkunen, first obtained by Reuters.

The timing matters. From August 2, 2026, the AI Act's transparency rules require clear labeling of content that meets the law's definition of a deepfake. Retailers want product imagery carved out before that date, and they have scale on their side: Zalando says 90 percent of the marketing content on its platform is now AI-generated. The company's content chief, Matthias Haase, framed the shift as moving from 'a planning mindset to a reacting one,' cutting production from weeks to days.

The word 'deepfake' is doing too much work

The real snag is definitional. 'Deepfake' was coined for nonconsensual fake pornography and fraud — impersonating a real person to deceive. Stretching the same word to cover a synthetic photo of a couch feels like a category error, and that's the retailers' opening: if the law's definition is fuzzy, they can argue their ads fall outside it. But the flip side is just as real. A generated product image can quietly misrepresent color, scale, fabric or finish, and a shopper has no way to know which pixels were invented.

This is the tension regulators have to resolve: write the rule too narrowly and most commercial AI imagery escapes it; write it too broadly and every retouched catalog photo carries a warning label that means nothing. Neither extreme helps the person actually clicking 'buy.'

Why it matters for you

Strip out the lobbying and this is about whether you can trust what you see in an online store. As AI-generated marketing goes from novelty to the default — 90 percent on one major platform already — the question of what gets labeled stops being abstract. If product ads are exempted, the line between a photo of a real thing and a convincing render of an imagined one disappears from your screen, and you're left guessing.

My take: the useful habit here isn't outrage, it's a small dose of skepticism. Treat a too-perfect product image the way you already treat a five-star review with no detail — assume it may be staged or generated, and lean on returns policies, real customer photos and specs over the hero shot. The label fight will play out in Brussels; your defense is the same one that always worked, just applied a little harder.

What I'd actually do

When an online product image looks flawless, scroll past it to the customer photos and the written specs before deciding. Those are far harder to fake than a hero shot — and until the labeling rules settle, they're the most reliable signal of what you're actually buying.

#AI regulation#EU AI Act#deepfakes

Related guides

EAEvgenii Arsentev

Author

Evgenii Arsentev

PhD · Chief Product Officer at a tech company

Want to actually build this?

Guides explain. The free course transforms — personalized, gamified, and built to get you shipping fast.

◉ Start the free course

Source: the-decoder.com