Agriculturally driven changes in soil phosphorus (P) are known to have persistent effects on local ecosystem structure and function, but regional patterns of soil P recovery following cessation of agriculture are less well understood. We synthesized data from 94 published studies to assess evidence of these land-use legacies throughout the world by comparing soil labile and total P content in abandoned agricultural areas to that of reference ecosystems or sites remaining in agriculture. Our meta-analysis shows that soil P content was typically elevated after abandonment compared to reference levels, but reduced compared to soils that remained under agriculture. There were more pronounced differences in the legacies of past agriculture on soil P across regions than between the types of land use practiced prior to abandonment (cropland, pasture, or forage grassland). However, consistent patterns of soil P enrichment or depletion according to soil order and types of post-agricultural vegetation suggest that these factors may mediate agricultural legacies on soil P. We also used mixed effects models to examine the role of multiple variables on soil P recovery following agriculture. Time since cessation of agriculture was highly influential on soil P legacies, with clear reductions in the degree of labile and total P enrichment relative to reference ecosystems over time. Soil characteristics (clay content and pH) were strongly related to changes in labile P compared to reference sites, but these were relatively unimportant for total P. The duration of past agricultural use and climate were weakly related to changes in total P only. Our finding of reductions in the degree of soil P alteration over time relative to reference conditions reveals the potential to mitigate these land-use legacies in some soils. Better ability to predict dynamics of soil nutrient recovery after termination of agricultural use is essential to ecosystem management following land-use change.