Objective Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by persistent deficits in social communication and restricted, repetitive behaviors. Increasing evidence suggests that these clinical manifestations are associated with atypical structural maturation of the brain""s triple-network system, including the default mode network (DMN), central executive network (CEN), and salience network (SN). However, how these networks exhibit distinct developmental trajectories remains insufficiently understood. In this study, we aimed to investigate network-specific structural maturation patterns in ASD using a BrainAGE framework based on voxel-wise gray matter density.Methods Structural magnetic resonance imaging (sMRI) data from 135 typically developing controls (TC) and 154 individuals with ASD were obtained from the Autism Brain Imaging Data Exchange (ABIDE) database. Voxel-based morphometry (VBM) was used to derive whole-brain gray matter density maps. Network-specific features were extracted from DMN, CEN, and SN templates. Feature selection was performed exclusively in the TC group using Pearson correlation analysis between voxel-wise gray matter density and chronological age, and voxels with P<0.001 were retained as age-related features. The ASD group used the corresponding voxel locations selected from the TC group to ensure identical feature space. Brain age prediction was performed using an Individual-Level Weighted Multi-Layer Perceptron (ILWMLP) BrainAGE model. The model was trained and internally validated in the TC group using leave-one-out cross-validation (LOOCV), and subsequently applied to the ASD group to estimate network-specific brain age. BrainAGE (delta age) was calculated as the difference between predicted brain age and chronological age. Group differences in developmental trajectories were assessed using Fisher Z test. A two-way mixed analysis of variance (ANOVA) was further conducted to compare atypical developmental patterns across networks. In addition, general linear model (GLM) analysis with age as a continuous variable was performed to validate the robustness of the findings.Results The BrainAGE model showed significant predictive performance in all three networks (DMN: r=0.600; CEN: r=0.778; SN: r=0.742). Compared with TC participants, individuals with ASD exhibited significantly altered age-dependent brain maturation patterns across all networks. Fisher Z test revealed significant group differences in developmental trajectories within DMN, CEN, and SN (all P<0.05). Importantly, the three networks demonstrated distinct developmental profiles. The DMN showed persistent delayed maturation across development, with the greatest deviation during adolescence. The CEN exhibited a dynamic pattern, shifting from delayed maturation in early development to relatively accelerated maturation in later stages. The SN showed progressively increasing delayed maturation with age. Two-way mixed ANOVA further confirmed significant differences in atypical developmental patterns among networks. GLM analysis yielded consistent results, supporting the robustness of the observed effects.Conclusion This study demonstrates that structural maturation abnormalities in ASD are not globally uniform but exhibit strong network specificity within the triple-network system. Although the DMN, CEN, and SN are functionally integrated, each network follows a distinct developmental trajectory during brain maturation. These findings extend the BrainAGE framework from global brain aging to network-level structural development and provide novel evidence for heterogeneous neurodevelopmental alterations in ASD. This network-specific perspective may contribute to the identification of imaging biomarkers and improve understanding of individualized brain maturation patterns in neurodevelopmental disorders.