Most weeds in arable fields are annuals that depend on seeds for survival over winter and for continued presence over years. Removing newly produced seeds would interrupt this cycle and constitutes an interesting alternative way of controlling weeds. Many animals in arable fields, such as carabid beetles, field crickets, and granivorous species of birds and mice, feed on weed seeds on the soil surface (seed predation). This way, seed-eating animals can remove considerable amounts of weed seeds. However, it is unclear whether this results in lower weed densities in subsequent years. For example, seed losses could be compensated by reduced seedling survival or enhanced seed production by the remaining weeds. Furthermore, there are usually large amounts of seeds already in the soil (seed bank), buffering any seed losses. In the current project, we investigated whether seed losses caused by seed-eating animals help controlling weed populations in agroecosystems by investigating the ability of weed populations to compensate for losses of newly produced seeds, or to buffer losses via the seed bank. We used a method that looked at each step in the life-cycle of a weed separately and another that investigated the overall effect of seed losses in one go. We found that the weed Echinochloa crus-galli (‘Hühnerhirse’, barnyard grass) in continuous maize fields could compensate via enhanced seedling survival, but not via reduced dormancy of seed mortality in the soil. Overall, barnyard grass in maize appeared considerably ‘seed limited’; i.e. most weed seeds that were added to the fields developed successfully into weed plants. They were much more seed-limited than plant populations in natural ecosystems. Seed banks mostly did not prevent successful establishment of weed plants, but delayed it. Therefore, weed populations were poorly able to compensate for or buffer seed losses caused by seed-eating animals. In conclusion, this project shows that the effect of seed predation is genuine, contributes to weed control and seed predation should be qualified an ecosystem service that is worthwhile preserving and enhancing. Observed levels of seed predation, between 38 and 70%, indicate that it already contributes to a significant reduction in weed population growth. These results are relevant to all farmers. The largest threat is losing residential granivores, and the ecosystem service they provide, due to ignorance and inexpert management. Followup research should therefore focus on the factors that limit the size, activity and composition of granivorous populations in arable fields.