Riser reactors are frequently applied in catalytic processes involving rapid catalyst deactivation. Typically heterogeneous flow structures prevail because of the clustering of particles, which impacts the quality of the gas–solid contact. This phenomenon results as a competition between fluid–particle interaction (i.e., drag) and particle–particle interaction (i.e., collisions). In this study, five drag force correlations were used in a combined computational fluid dynamics–discrete element method Immersed Boundary Model to predict the clustering. The simulation results were compared with experimental data obtained from a pseudo-2D riser in the fast fluidization regime. The clusters were detected on the basis of a core–wake approach using constant thresholds. Although good predictions for the global (solids volume fraction and mass flux) variables and cluster (spatial distribution, size, and number of clusters) variables were obtained with two of the approaches in most of the simulations, all the correlations show significant deviations in the onset of a pneumatic transport regime. However, the correlations of Felice ( Int. J. Multiphase Flow 1994, 20, 153−159) and Tang et al. [ AIChE J. 2015, 61 (2), 688−698] show the closest correspondence for the time-averaged quantities and the clustering behavior in the fast fluidization regime.