Temporal trends in attributes of restored ecosystems have been described conceptually as restoration trajectories. Measures describing the maturity or ecological integrity of a restoration site are often assumed to follow monotonically increasing trajectories over time and to eventually reach an asymptote representative of a reference ecosystem. This assumption of simple, predictable restoration trajectories underpins federal and state policies in the United States that mandate wetland restoration as compensation for wetlands damaged during development. We evaluated the validity of this assumption by tracking changes in 11 indicators of floristic integrity, often used to determine legal compliance, in 29 mitigation wetlands. Each indicator was expressed as a percentile relative to the distribution of that indicator among >100 naturally occurring reference wetlands. Nonlinear regression was used to fit two alternative restoration trajectories to data from each site: an asymptotic (negative exponential) increase in the indicator over time and a peaked (double exponential) relationship. Depending on the particular indicator, between 48% and 76% of sites displayed trends that were at least moderately well described (R² > 0.5) by one of the two models. Floristic indicators based on species richness, including native richness, number of native genera, and the floristic quality index, rapidly increased to asymptotes exceeding levels in a majority of reference wetlands. In contrast, indicators based on species composition, including mean coefficient of conservatism and relative importance of perennial species, increased very slowly. Thus, some indicators of restoration progress followed increasing trajectories and achieved or surpassed levels equivalent to high-quality reference sites within five years, whereas others appeared destined to either not reach equivalency or to take much longer than mitigation wetlands are typically monitored. Finally, some indicators of restoration progress, such as relative importance of native species, often increased over the first five to 10 years and then declined, which would result in a misleading assessment of progress if based on typical time scales of monitoring. Therefore, the assumption of simple, rapid, and predictable restoration trajectories that underlies wetland mitigation policy is unrealistic.