Agroforestry (AFS), the integration of trees into farming systems, offers a promising strategy to enhance agroecosystem functions and address climate change. Its strength lies in the capacity of trees to exploit both surface and subsurface resources beyond the reach of crops, while also contributing to the rehabilitation of degraded farmland through increased above- and below-ground organic matter inputs. However, the observed effects of AFS on these functions are known to vary depending on pedoclimatic conditions and the specific types of AFS implemented and specific soil-related process considered. Here, we detailed the effects of AFS compared to treeless agricultural systems on a global scale based on a second-order meta-analysis, synthesizing findings from 1’500 primary studies and 26 relevant meta-analyses identified through Web of Science, Scopus, and prior systematic reviews. Each meta-analysis was evaluated using standardized quality criteria, and the overlap of primary studies was analysed to assess the independence of the evidence base. The MA quality scores ranged from 25% to 100%, with over half of the MAs having a score of 75% or higher. However, half of the MAs didn’t weight individual effect sizes or perform any heterogeneity or publication bias analyses. We developed a knowledge gap map for soil and water-related outcomes, highlighting well-covered topics in the literature and identifying areas needing further research. Our findings indicate that AFS generally improves soil carbon (28% [19-37% 95%CI]), and enhances both chemical (37% [17-59% 95%CI]) and biological soil quality (59% [1-152% 95%CI]). Soil carbon parameters (SOC stocks and content) are increased regardless of practice type and pedoclimate, except in arid pedoclimates where soil carbon is estimated to increase by up to 77% [24-121% 95%CI] %), though data paucity in these areas limits the reliability of these estimates. Improvements in physical soil quality (37%, [8-74% 95%CI]), erosion control (183%, [53-423% 95%CI]), and water retention (73%, [63-83% 95%CI]) were also observed, albeit with less robust evidence due to effect heterogeneity and limited meta-analytic coverage. While the broad scope of this second-order meta-analysis restricts conclusions to aggregated indicators rather than specific soil processes, it offers a global quantitative perspective on the benefits of AFS for soil health and water conditions. This work also assesses the reliability of existing meta-analytic evidence and provides a basis for scaling up AFS practices, supporting informed decision-making for the implementation of different AFS practices across different environmental contexts.