Surprising users in UX design backfires because it violates the Principle of Least Astonishment: interfaces should not require users to stop and figure out unexpected behavior. When a feature does not match the expectations set by its visual design or labeling, users experience friction, confusion, and frustration rather than delight.
Users form expectations based on established design patterns and conventions across the entire web. When those conventions are broken (a star rating system where one star is the best rather than the worst, auto-playing audio on a content page, a hover interaction that launches an unwanted video), users must abandon their prior mental model and learn a new one for that specific site. Each surprise interrupts the user's flow, wasting mental effort that was directed toward completing a task. Netflix's autoplay feature demonstrates the pattern: users trying to browse show details had to navigate around hover areas that triggered loud trailers, turning a browsing task into an avoidance task. The key principle is that function must match the expectations set by messaging and visual design. Redesigning known patterns for the sake of differentiation forces users to relearn interactions they have already mastered, and the cognitive cost of that relearning almost always outweighs any branding benefit.