While interpretation-bias modification (IBM) is an effective intervention for treating anxiety, it is not broadly used in clinical or daily practice. To this end, this study developed and tested a smartphone-based IBM application. We adopted the ambiguous situation paradigm as an intervention task in conjunction with robust training materials that broadly covered situations encountered in daily life. We recruited participants with high-trait anxiety and divided them into three groups: (1) positive training; (2) 50% positive-50% negative training; and (3) no-training control. The first two groups completed 28 days of smartphone-based training (IBM in positive cases), and all groups completed six rounds of assessments. The smartphone-based IBM training changed positive and negative endorsements and more specific measures of interpretation bias, thus reducing anxiety. The results also showed that changes in the number of negative interpretations played a mediating role in anxiety reduction. It is notable that the attrition rate was extremely low across the experiment. Our follow-up showed that positive gains persisted throughout the intervening period. Smartphone-based IBM can help individuals with anxiety shift negative biases, broaden their thoughts, enhance their information processing, and effectively target the clinical features of anxiety.