Xia-Gibbs syndrome (XGS) is a rare neurodevelopmental disorder with considerable clinical heterogeneity. To further characterize the syndrome’s heterogeneity, we applied latent class analysis (LCA) on reported cases to identify phenotypic subtypes. By searching PubMed, Embase, China National Knowledge Infrastructure and Wanfang databases from inception to February 2024, we enrolled 97 cases with nonsense, frameshift or missense variants in the AHDC1 gene. LCA was based on the following 6 phenotypes with moderate occurrence and low missingness: ataxia, seizure, autism, sleep apnea, short stature and scoliosis. After excluding cases with missing data on all LCA variables or with unmatched phenotype-genotype information, a total of 85 cases were selected for LCA. Models with 1–5 classes were compared based on Akaike Information Criterion, Bayesian Information Criterion, Sample-Size Adjusted BIC and entropy. We used multinomial logistic regression (MLR) analyses to investigate the phenotype-genotype association and potential predictors for class membership. LCA revealed 3 distinct classes labeled as Ataxia subtype (n = 11 [12.9%]), Sleep apnea & short stature subtype (n = 23 [27.1%]) and Neuropsychological subtype (n = 51 [60.0%]). The commonest Neuropsychological subtype was characterized by high estimated probabilities of seizure, ataxia and autism. By adjusting for sex, age and variant type, MLR showed no significant association between phenotypic subtype and variant position. Age and variant type were identified as predictors of class membership. The findings of this review offer novel insights for different presentations of XGS. It is possible to deliver targeted monitoring and treatment for each subtype in the early stage.