Dark Patterns, AI and Regulation: DSA, DMA and Consumer Law

By the UXreviewer.io team4 min read

Dark patterns are interface designs that push users into decisions they would not otherwise make, such as hidden costs, hard-to-cancel subscriptions or pre-selected consent. In the EU they are addressed by several overlapping rules: consumer protection law, the GDPR, the Digital Services Act, the Digital Markets Act and, for AI-driven manipulation causing significant harm, the AI Act. AI raises the stakes because it can tailor pressure to each individual in real time.

This article maps the rules to concrete design patterns, explains what changes when agents and personalisation enter the picture, and ends with an audit checklist. It is a design-focused overview, not legal advice.

Which EU rules cover dark patterns?

The European Commission has also been preparing a Digital Fairness initiative that may add specific rules on dark patterns, addictive design and personalisation. Check its current status before planning long-term changes.

• Unfair Commercial Practices Directive: prohibits misleading and aggressive commercial practices toward consumers, which covers many dark patterns such as fake urgency, hidden information and nagging. It applies across sectors and company sizes.
• Consumer Rights Directive: among other things, prohibits pre-ticked boxes for additional payments and requires clear information before purchase. Amendments add a withdrawal function for online distance contracts, applicable from 19 June 2026.
• GDPR: consent must be freely given, specific, informed and unambiguous, which rules out pre-ticked boxes and designs that make refusal much harder than acceptance.
• Digital Services Act, Article 25: online platforms must not design interfaces that deceive, manipulate or otherwise materially impair users' ability to make free and informed decisions, where those practices are not already covered by consumer law or the GDPR.
• Digital Markets Act: gatekeepers may not use interface design to circumvent their obligations, for example by making choices or data portability unnecessarily difficult.
• AI Act, Article 5: prohibits AI systems using manipulative or deceptive techniques, or exploiting vulnerabilities, that materially distort behaviour and cause or are likely to cause significant harm, applicable since 2 February 2025.

How does AI change the dark pattern problem?

Classic dark patterns are static: the same confusing cancellation flow for everyone. AI makes persuasion adaptive. A model can learn which message makes a specific user hesitate, choose the moment they are most likely to accept, and phrase pressure in a friendly conversational tone that feels like advice rather than marketing.

Conversational agents add new forms: a support bot that repeatedly redirects cancellation requests to retention offers, a companion app that expresses sadness when the user wants to leave, or a shopping agent that quietly favours higher-margin products while presenting itself as neutral. Each can fall under consumer law even when no single screen looks manipulative.

Which patterns deserve the closest scrutiny?

In UX reviews, these patterns most often carry regulatory risk, with or without AI:

• Asymmetric choices: 'Accept all' prominent, 'Reject' hidden or several clicks deeper.
• Roach motels: easy sign-up, difficult cancellation, especially via chat-only or phone-only routes.
• False urgency and scarcity: countdowns and stock warnings not based on real data, including AI-generated ones.
• Confirmshaming: guilt-inducing wording on the decline option, now sometimes generated per user.
• Drip pricing: fees revealed late in checkout.
• Disguised advertising: sponsored recommendations from an assistant without clear labelling.
• Nagging: repeated prompts until the user gives in.

What does fair design look like for AI agents?

Fair agents act in the user's interest within clearly stated limits. They disclose when recommendations are sponsored or limited to a catalogue, they accept a clear 'no' without repeated persuasion, and they complete cancellation or opt-out requests as directly as purchases.

Practical rules of thumb: make the agent's goal transparent; never let retention tactics block a stated intention to cancel; apply the same effort to declining as to accepting; and avoid emotional pressure in persona design. Our article on agentic commerce checkout UX shows how this works in buying flows, and our piece on AI ethics in product design covers the broader principles.

How do you audit your product for dark patterns?

An external reviewer helps, because internal teams often see their own retention flows as 'just good marketing'. A psychology-informed UX audit names the mechanism behind each pattern, which makes the case for change much easier.

1. List decision points: sign-up, consent, upgrades, add-ons, checkout, cancellation, data sharing, notifications.
2. For each, compare the effort and visibility of the 'yes' path and the 'no' path.
3. Check that urgency, scarcity, social proof and price claims are true and verifiable.
4. Review AI-generated messages and agent behaviour across many user profiles, not one demo account.
5. Test cancellation and opt-out through every channel, including chat and agent.
6. Record findings, fixes and decisions as evidence for regulators and partners.

Is ethical design bad for conversion?

Short-term metrics can drop when a manipulative pattern is removed. Over a longer horizon, trust, lower refund and chargeback rates, fewer complaints and lower regulatory exposure usually matter more. Many teams find that clearer flows also convert better for users who actually want the product.

Measure what matters: retention of users who chose freely, satisfaction after purchase, and support contacts, alongside conversion. Our article on UX metrics that matter suggests a balanced set.

Frequently asked questions