Dark patterns are design choices that lead users to do something they would not do with complete information and free choice: difficult account deletion, hidden pre-selected costs, fake urgency, confirmshaming. With AI, these patterns become more personalised — and therefore more invisible.
Manipulative personalisation exploits behavioural data to calibrate pressure: "Only 2 seats left" shown only to anxious users, discounts shown to those more likely to buy immediately, opaque dynamic pricing that maximises willingness-to-pay. The designer who implements these flows participates in unfair practices.
AI-generated artificial urgency — "personalised" notifications that simulate scarcity or FOMO — erodes long-term trust for short-term metrics. Ask yourself: is this urgency verifiable? Can the user control it? Or is it manufactured by the algorithm?
Algorithmic nudging in health, finance or information can steer high-impact choices without the user noticing. "We suggest this plan" on an opaque basis is not assistance — it is persuasion. The line between help and manipulation runs through transparency and reversibility.
As a designer you can refuse: document the pattern as risky, propose alternatives (verifiable countdowns, fixed prices, explainable recommendations), cite regulations (Digital Services Act, Unfair Commercial Practices). Sometimes showing a negative user scenario in a test is enough to change the PM's mind.
The deceptive.design community and FTC/EU guidelines are converging towards sanctions for dark patterns. AI makes them more effective — but not more acceptable. On the contrary, it increases the responsibility of those who design them.