Abstract Background Telecommunications fraud causes financial and psychological harm, yet how individuals who experienced fraud retrospectively describe progression from initial trust to emotional escalation and reduced behavioral control remains unclear. Objective This study aimed to develop a stage-based psychological framework for telecommunications fraud and examine whether person-level cross-sectional associations provide quantitative elaboration. Methods Study 1 used grounded-theory analysis of semistructured interviews with 71 individuals who experienced telecommunications fraud; 46 interviews informed model development and 25 assessed theoretical saturation. Study 2 analyzed 4722 university-based survey responses from Chongqing, including 409 students with confirmed fraud experience. A balanced case-control sample comprised all 409 participants with confirmed fraud experience and 409 randomly selected participants without reported fraud experience. Two-tailed independent-samples t tests, adjusted logistic regression, and 10 nonoverlapping resampling iterations examined correlates of victimization. Among defrauded people, 2 theory-specified PROCESS Model 6 analyses with 5000 bootstrap resamples estimated cross-sectional indirect associations involving Truth-Seeking or Cognitive Maturity, Gullibility, Difficulties in Emotion Regulation, and Brief Self-Control Scale (BSCS) scores. Results Study 1 yielded a 3-stage interpretive framework of retrospective narratives: Credulity Priming, Affective Manipulation, and Behavioral Dyscontrol. Desire and fear loops, as well as overexpectation events, were characterized by a narrated escalation. In study 2, defrauded people had lower Truth-Seeking and Cognitive Maturity and higher Gullibility, Difficulties in Emotion Regulation, and BSCS scores (indicating poorer self-control) than sampled nondefrauded people. In adjusted analyses, higher Gullibility was associated with higher odds of victimization, whereas higher Truth-Seeking was associated with lower odds; the interpretation of these adjusted correlates was based primarily on their consistent reproduction in the student-only sensitivity models. Among defrauded people, both PROCESS models yielded statistically significant model-specified cross-sectional indirect association estimates, but concurrent postevent measurement precluded temporal or causal inference. Across studies, convergence was the strongest for trust-related appraisal; evidence for affective and behavioral domains was complementary rather than confirmatory. Conclusions Retrospective narratives supported a 3-stage interpretive framework, while cross-sectional quantitative findings most consistently linked lower Truth-Seeking and higher Gullibility to victimization status in this sample. These findings motivate prospective evaluation of stage-sensitive prevention strategies; the proposed temporal sequence and intervention implications require direct testing in more diverse populations.