Cover crops are considered to be beneficial for multiple ecosystem services, and they have been widely promoted through the Common Agricultural Policy (CAP) in the EU and Farm Bill Conservation Title Programs, such as the Environmental Quality Incentives Program (EQIP), in the USA. However, it can be difficult to decide whether the beneficial effects of cover crops on some ecosystem services are likely to outweigh their harmful effects on other services, and thus to decide whether they should be promoted by agricultural policy in specific situations. We used meta-analysis to quantify the effects of cover crops on five ecosystem services (food production, climate regulation, soil and water regulation, and weed control) in arable farmland in California and the Mediterranean, based on 326 experiments reported in 57 publications. In plots with cover crops, there was 13% less water, 9% more organic matter and 41% more microbial biomass in the soil, 27% fewer weeds, and 15% higher carbon dioxide emissions (but also more carbon stored in soil organic matter), compared to control plots with bare soils or winter fallows. Cash crop yields were 16% higher in plots that had legumes as cover crops (compared to controls) but 7% lower in plots that had non-legumes as cover crops. Soil nitrogen content was 41% lower, and nitrate leaching was 53% lower, in plots that had non-legume cover crops (compared to controls) but not significantly different in plots that had legumes. We did not find enough data to quantify the effects of cover crops on biodiversity conservation, pollination, or pest regulation. These gaps in the evidence need to be closed if cover crops continue to be widely promoted. We suggest that this novel combination of multiple meta-analyses for multiple ecosystem services could be used to support multi-criteria decision making about agri-environmental policy.