Companion plant intercropping involves growing a cash crop with another plant that is not harvested, to confer a set of benefits on the crop and the environment. Weed control and herbicide use reduction are one of the principal reasons for adopting this approach. Companion plants should compete with weeds for light, nutrients and water, but they may also compete with the crop. The species grown must therefore be carefully chosen and managed so as to outcompete weeds but limit competition with the crop and yield loss. In this meta-analysis, we aimed to quantify the effects of companion plants on weed regulation and cash crop yields, and to analyze their sources of variability. We reviewed different intercropping systems involving an annual cash crop and a legume companion plant from around the world. We report data from 34 scientific articles, corresponding to 476 experimental units (i.e. combinations of site × year × cash crop × legume companion plant species × agricultural practices), and we explore whether intercropping with legume companion plants can control weeds while maintaining crop yield. Yield and weed biomass ratios were analyzed as response variables. We used the type of cash crop (straw cereals, maize or other crops), the methods used to establish the companion plants (living mulch, synchronized sowing or relay intercropping) and the overlap between the growth periods of the companion plants and the cash crop as explanatory variables. Intercropping with a companion plant resulted in a lower weed biomass and a higher yield (win-win situation) than non-weeded or weeded control treatments, in 52% and 36%, respectively, of the experimental units considered. A higher weed biomass associated with a lower yield (lose-lose) was observed in only 13% and 26% of the experimental units, in comparisons with non-weeded and weeded control treatments, respectively. Considering all the experimental units together, the companion plants had no significant effect on cash crop yield, but significantly decreased weed biomass, by 56% relatively to a non-weeded control treatment, and 42% relative to a weeded control treatment. The greatest benefits from companion plant intercropping were reported for maize, with yields 37% higher than those for non-weeded control treatments. The other explanatory variables tested had no significant effect on yield or weed control. Thus, the use of legume companion plants generally seems to enhance weed control without reducing crop yield, but the conditions giving rise to win-win situations should be explored further, to encourage the spread of this technique among farmers.