Dispersal is one of the most important drivers of community assembly. Understanding how dispersal impacts spatial variations in community composition (beta diversity) is crucial for predicting biodiversity change during the Anthropocene. Classic theories and a large amount of empirical evidence have led to a common belief that increasing dispersal rate should lead to regional community homogenization (lower beta diversity). However, recent empirical studies show that increasing dispersal can also lead to regional community divergence when interacting with other processes such as disturbance and priority effect. The underappreciation of the complex relationship between beta diversity and dispersal might impede our ability to make accurate ecological forecasts. To address this concern, I examine the beta diversity–dispersal relationship through extinction and colonization dynamics in meta‐community models. I find that a wide range of beta diversity–dispersal relationships (e.g. negative, positive, hump‐shaped) can be generated by changes in species‐level occurrence probabilities when meta‐population dispersal (rather than mainland–island dispersal), disturbance and strong biotic interactions are at play. Specifically, reducing the variance of species‐level occurrence probabilities leads to community divergence and increasing the mean occurrence probability leads to community homogenization; a positive or hump‐shaped beta diversity–dispersal relationship emerges when the effect of the decreasing variance is stronger than the effect of increasing mean. These findings highlight the intrinsic constraint of occupancy stochasticity on beta diversity changes and echo the call for more realistic manipulations of stochastic dispersal and a detailed understanding of the interaction networks to test various beta diversity–dispersal relationships in future empirical studies.