The use of AI in website design is becoming more common, and with thatcomes growing concern about how it might shape user experiences, sometimesin ways that aren’t immediately noticeable. A specific area of interest in thisstudy are dark patterns, which are design tricks that can push users towardcertain actions, often without them realizing it. The idea was to take a closerlook at how AI could change or adapt these patterns depending on how someoneinteracts with a website in real time.

This study utilizes an A/B testing methodology on a simulated e-commercewebsite, where one version included statically implemented dark patterns, andthe other version included dark patterns dynamically implemented by AI. TheAI system monitored real-time behavior, such as mouse movement and clickpatterns, and responded by selecting different manipulative design elementsdepending on the user's behavior. In addition to the experiment, we conductedstructured interviews to gain a better understanding of how participantsexperienced the two versions.

The results of the study show that AI-based personalization often made the darkpatterns more effective, even when fewer were shown. This raised importantethical questions around user autonomy, transparency, and consent. Whilepersonalization can be useful, the line between helpful and manipulative designis not always easy to draw, especially when that line is shaped by AI.