Soil fauna play a key role in the soil food web and is an important index to evaluate the soil environment and ecosystem. Crop straw residue is recommended for application to agricultural fields to improve soil quality. However, the effects of continuous straw return on the soil faunal community remain unclear, and knowledge regarding purple soil is limited. Using the conserved cytochrome oxidase I (COI) gene fragment of eukaryotic cells as a molecular marker, we examined the responses of the soil faunal community to eighteen-year continuous straw return in upland purple soil of southwestern China, i.e., RSDN (crop residues plus synthetic nitrogen), RSDNP (crop residues returned with nitrogen and phosphorus) and RSDNPK (crop residues returned with nitrogen, phosphorus and potassium), and treatments without straw, i.e., N (synthetic N fertilizer) and NPK (synthetic fertilizer: nitrogen, phosphorus and potassium). Soil nematodes were also examined using traditional morphological methods, and the accuracy of the morphological and molecular methods in characterizing soil nematodes was compared. Our results showed that different fertilization treatments significantly changed the soil faunal community structure (Adonis test, R2 = 0.43, p = 0.011). According to LEfSe (linear discriminant analysis effect size), Lymnaeidae, Tetragnathidae, and Pyramidulidae et al. were associated with straw return, and Coryne, Corynidae, and Anthoathecata et al. were associated with phosphate addition. Diplopoda was associated with both straw and phosphate amendments. Straw return significantly increased the Margalef richness index (D) and species number (S), which were closely related to soil organic matter (SOM) and total phosphorus (TP) (p < 0.05). Structural equation model (SEM) analysis revealed that straw and phosphorus addition promoted the diversity and species number of soil fauna by changing soil SOM and AP (available phosphorus), respectively. These results highlighted that SOM and P were important factors that changed the soil faunal community and promoted diversity. Additionally, we found that both the traditional morphological approach and molecular method based on the COI gene had biases in characterizing the soil nematode community.