BACKGROUND: Emerging evidence suggests that the atherogenic index of plasma (AIP), a simple marker of lipid-related metabolic risk, may relate to dysglycemia in a nonlinear fashion. However, whether this association differs across body mass index (BMI) categories, particularly in terms of nonlinear risk patterns and thresholds, remains unclear. METHODS: We analyzed data from 100,473 Chinese adults who underwent baseline health examinations and had baseline FPG < 5.6 mmol/L. Because HbA1c and OGTT data were unavailable, the outcome was defined as FPG-defined incident prediabetes (5.6-6.9 mmol/L). Participants were classified into underweight, normal weight, overweight, and obesity groups according to Chinese BMI criteria. Restricted cubic spline (RCS) and Cox regression were used to evaluate BMI-specific associations between AIP and prediabetes risk and to test for interaction by BMI. RESULTS: During a median follow-up of 3.0 years, 12,371 participants developed FPG-defined incident prediabetes. A significant interaction between AIP and BMI was observed (P for interaction < 0.001). BMI-stratified analyses showed that AIP was associated with FPG-defined prediabetes risk only among normal-weight and overweight participants, but not among underweight or obese participants. RCS analyses further identified nonlinear associations in these two BMI groups, with model-derived inflection points at AIP = -0.20 in normal-weight individuals and AIP = 0.057 in overweight individuals. CONCLUSION: BMI modified the association between AIP and FPG-defined prediabetes risk. Significant nonlinear and threshold-dependent associations were observed only in normal-weight and overweight Chinese adults.