Intelligent digital services, arising from the deep integration of digital and intelligent technologies into the service sector, have enhanced service efficiency and personalization while simultaneously triggering a trust crisis among users. This study introduces the concept of a trust deficit into the domain of intelligent digital services to characterize situations in which displacement of human–AI trust lead to persistently lower service trust levels than expected. As a key constraint on the effectiveness of intelligent digital services, trust deficits cannot be adequately explained by single-dimensional frameworks grounded solely in interpersonal trust or technical trust. This research focuses on the multidimensional structure and interrelationships of trust deficits, proposing a four-dimensional generative model encompassing ability, benevolence, integrity, and controllability. It uncovers the evolutionary pathways of trust deficits characterized by cross-cycle amplification, multidimensional coupling, and heterogeneous decay, and examines the mechanisms through which service type, contextual risk, and users’ intelligent digital literacy influence their evolution. Scenario simulations reveal that, following service failure, tool-based and empathy-based intelligent digital services differ markedly in their dominant deficit dimensions and evolutionary trajectories; service risk significantly accelerates the growth of trust deficits and reshapes their evolutionary patterns; and higher intelligent digital literacy effectively mitigates the escalation of trust deficits, even enabling their convergence to zero over multiple service interactions. This study broadens the theoretical boundaries of trust research in the context of intelligent digital services and offers a conceptual foundation for developing strategies to manage trust deficits.