Understanding susceptibility to misinformation is essential for curbing its spread and making effective interventions. However, traditional misinformation susceptibility test (MIST) is often constrained by high costs, extended durations, and ethical concerns. To overcome these limitations, we introduce an agent-based simulation framework powered by Large Language Models (LLMs) to examine the influence of demographic and psychological factors on information credibility judgments within the Chinese context. Leveraging the DeepSeek model, we construct two participant-agent profiles incorporating demographic attributes, Big Five personality traits and individual profiles. The framework simulates individual credibility assessments from three psychological mechanisms: stereotype, cognitive consistency, and conformity, while Ordinary Least Squares (OLS) regression and the Explainable Boosting Machine (EBM) are employed to interpret the simulation outputs and identify key predictors. Results show that age is positively associated with susceptibility to misinformation in public domains, while higher education consistently mitigates susceptibility. Among personality traits, agreeableness is the most consistent positive predictor of credibility judgments. The EBM results further highlight that sentiment polarization and cognitive consistency are the key factors in misinformation detection. Our approach presents a scalable and cost-effective alternative to traditional MIST methods, offering significant implications for misinformation intervention and the modeling of cognitive processes.