Background/Question/Methods . Traditional approaches and metrics to assess ecosystem recovery focus on simple attributes such as taxonomic richness or carbon accumulation in soils. Assessments of the restoration performance worldwide show that, when traditional restoration guidelines based on the recovery of those simple metrics are followed, restored ecosystems may only recover part of their lost biodiversity, functions, and benefits to societies, even after decades or centuries. They also show that active restoration efforts may not yield better results than naturally regenerating ecosystems. This suggests that traditional approaches may be simplified abstractions to achieve a sustained recovery. We address this limitation by focusing on recovering ecosystems to understand how complexity recovers after anthropogenic disturbance over long time-periods (centuries or more) with two approaches, meta-analysis and empirical observations in space-for-time substitutions. This allows us to respond to two key questions, how long is ecosystem recovery? And, what are the key mechanisms that regulate the recovery of ecosystem structure and function? We have used multiple meta-analytical approaches (response ratios, recovery debt, chronosequences) in all kinds of ecosystems globally. Empirically we have detected changes in complexity through time focusing on the interaction between the plant community and soil fungi in old mines in Spain or the ancient Norse farms of Greenland. Results/Conclusions . Results from the meta-analyses and empirical observation suggest that even simple ecosystem attributes (i.e., diversity or carbon and nitrogen cycling) may not recover after several centuries of recovery. Our meta-analysis of 104 forest chronosequences shows a >400 years to recover diversity, composition and nitrogen cycling. Results from the Norse farms show that after >800 years of abandonment subarctic plant communities differed between farmed and undisturbed areas, while the average species richness recovered. Fungal functional guilds were still dominated by pathogens and detritivores in former farms while undisturbed sites after that time were mutualists dominated undisturbed sites. Our results support the hypothesis that simple metrics may recover earlier than complex ones and the hypothesis that recovery promotes mutualistic interactions. Overall, complete ecosystem recovery after anthropogenic disturbance may be a centennial to millennial process, which must be accounted in environmental regulations and large-scale restoration strategies. Actions to accelerate recovery would range from restoring meta-community hub species that reduce vulnerability to favoring mutualistic soil fungi.