AI and consumer protection

Short answer: AI that misleads, pressures or exploits the weaknesses of consumers falls under the existing ban on unfair commercial practices (Directive 2005/29/EC) — and sometimes also under the prohibitions in the AI Act. Both regimes apply side by side, and consumer law applies regardless of whether your AI system is high-risk.

What consumer law prohibits

The Unfair Commercial Practices Directive bans two main categories: misleading practices (false or confusing information) and aggressive practices (harassment, coercion, undue influence). The test is whether the average consumer is led to a decision they would not otherwise have taken.

That touches AI directly. A recommendation system inventing scarcity, a chatbot making false claims, or an interface that deliberately makes opting out hard — these are classic unfair practices in new clothing.

Dark patterns and personalised deception

AI makes two things easier. First, dark patterns: deceptive interface designs that steer consumers towards a choice. Second, personalisation: tailored deception, matched to an individual's profile or moment of weakness.

That personalisation is especially dangerous, because the "average consumer" is no longer the benchmark — everyone gets their own optimised persuasion strategy. Consumer law accounts for vulnerable groups, and personalised pressure can more readily count as aggressive or misleading.

The boundary with the AI Act

The AI Act bans AI that uses manipulative or deceptive techniques to materially distort behaviour, and AI that exploits vulnerabilities (age, disability, social or economic situation). See prohibited AI practices.

That ban partly overlaps with consumer law but does not cover it: many unfair practices fall below the AI Act's high threshold yet are still prohibited as a commercial practice. In addition, Article 50 requires users to know they are dealing with AI and synthetic content to be recognisable — which helps prevent deception.

Fake content and trust

AI-generated reviews, testimonials and "independent" comparisons are misleading if presented as genuine. Displaying fabricated social proof is an unfair practice. This connects to AI and disinformation, where the same techniques operate in a political context.

What to do

• Test interfaces for steering or misleading elements — the design, not just the text.
• Label AI content: make chatbots, generated images and synthetic reviews recognisable.
• Be cautious with personalisation that targets weaknesses or invents urgency.
• Verify claims an AI system generates itself for accuracy and substantiation.
• Combine the regimes: consumer law and the AI Act, because the former applies outside high-risk too.

AI Act compliance says little about your marketing. The ban on unfair commercial practices applies independently of it.