The present study aimed to explore the latent structure of the 21-item Peters et al. Delusions Inventory (PDI-21) and to test the invariance of its factor structure over time and across samples. The PDI-21 was administered in two samples; one consisting of 1655 undergraduate students investigated in three waves, each separated by six months; and another consisting of 196 outpatients with schizophrenia. Exploratory factor analysis was performed to explore the internal structure of the PDI-21 based on number of beliefs and the grand total scores separately. The number of factors was determined by optimal implementation of parallel analysis. Confirmatory factor analysis, cross-time and cross-sample invariance analyses were carried out with Mplus. Both exploratory factor analysis and the optimal implementation of parallel analysis (based on the number of beliefs and the total score of the PDI-21) suggested a one-factor solution. However, the confirmatory factor analysis revealed a single dimension structure based on number of beliefs only, which exhibited goodness of fit and stability across time and samples. Our study demonstrated a single-dimension structure of the PDI-21, which can be widely used in screening the number of delusional ideations both in clinical and non-clinical populations.