Paleoecological analyses that test for spatial or temporal variation in diversity must consider not only sampling and preservation bias, but also the effects of temporal scale (i.e., time-averaging). The species-time relationship (STR) describes how species diversity increases with the elapsed time of observation, but its consequences for assessing the effects of time-averaging on diversity of fossil assemblages remain poorly explored. Here, we use a neutral, dispersal-limited model of metacommunity dynamics, with parameters estimated from living assemblages of 31 molluscan data sets, to model the effects of within-habitat time-averaging on the mean composition and multivariate dispersion of assemblages, on diversity at point (single station) and habitat scales (pooled multiple stations), and on beta diversity. We hold sample size constant in STRs to isolate the effects of time-averaging from sampling effects. With increasing within-habitat time-averaging, stochastic switching in the identity of species in living (dispersal-limited) assemblages (1) decreases the proportional abundance of abundant species, reducing the steepness of the rank-abundance distribution, and (2) increases the proportional richness of rare, temporally short-lived species that immigrate from the neutral metacommunity with many rare species. These two effects together (1) can shift the mean composition away from the non-averaged (dispersal-limited) assemblages toward averaged assemblages that are less limited by dispersal, resembling that of the metacommunity; (2) allow the point and habitat diversity to increase toward metacommunity diversity under a given sample size (i.e., the diversity in averaged assemblages is inflated relative to non-averaged assemblages); and (3) reduce beta diversity because species unique to individual stations become shared by other stations when limited by a larger but static species pool. Surprisingly, these scale-dependent changes occur at fixed sample sizes and can become significant after only a few decades or centuries of time-averaging, and are accomplished without invoking ecological succession, environmental changes, or selective postmortem preservation. Time-averaging results in less inflation of diversity at habitat than at point scales; paleoecological studies should thus analyze data at multiple spatial scales, including that of the habitat where multiple bulk samples have been pooled in order to minimize time-averaging effects. The diversity of assemblages that have accumulated over 1000 years at point and habitat scales is expected to be inflated by an average of 2.1 and 1.6, respectively. This degree of inflation is slightly higher than that observed in molluscan death assemblages at these same spatial scales (1.8 and 1.3). Thus, neutral metacommunity models provide useful quantitative constraints on directional but predictable effects of time-averaging. They provide minimal estimates for the rate of increase in diversity with time-averaging because they assume no change in environmental conditions and in the composition of the metacommunity within the window of averaging.