This study examines how different null models influence the analysis of phylogenetic diversity indices and their relationships with elevation in plant communities in Chongqing, China. We calculated SESPD, SESMPD, and SESMNTD using seven null models (Richness, Taxa.table, Frequency, Independent swap, Trialswap, Sample.pool, and Phylogeny.pool) based on data from 58 sampling sites. Results indicate that although ANOVA showed no significant differences in SESPD and SESMNTD among the null models, their elevation patterns and the variance explained by environmental factors in RDA varied greatly. For instance, the Frequency (Freq) model explained the highest variance (68%) in SESPD, while Trialswap (TriS) explained only 30%. Elevation patterns ranged from monotonic to bimodal relationships, depending on the null model used. Our findings emphasize the critical role of null model selection in studying phylogenetic diversity. We suggest comparing multiple null models to identify the most reliable one, as different models can lead to contrasting conclusions. This approach can enhance the robustness of ecological analyses and provide deeper insights into the processes shaping community structure.