Perceived Trustworthiness as a Pathway Linking ADHD Traits to Large Language Models Dependency: Evidence from UK and Chinese Samples

Large language models (LLMs) create contexts in which individual differences may shape reliance on digital systems. Drawing on neurodiversity, this study examined whether perceived trustworthiness of user’s primary LLM mediates associations between attention-deficit/hyperactivity disorder (ADHD) traits and LMM dependencyinstrumental and relationship LLM dependency. Adults in the UK (n = 567) and China (n = 577) reported ADHD traits, perceived LLM trustworthiness of LLM, and both types of dependency. Across samples, higher ADHD traits were associated with lower trustworthiness, whereas greater trustworthiness was associated with greater instrumental and relationship dependency, yielding negative indirect effects. In the UK, direct ADHD-dependency associations were non-significant after controlling for trust, and mediation produces a significant negative total effect for relationship dependency. In China, negative indirect effect coexisted with positive direct associations, creating opposing pathways whose total effects depends on their relative magnitude. Implications for inclusive LLM design without unintendend dependency and supporting neurodiverse users are discussed.