Some individuals (helpers) in cooperatively breeding species provide alloparental care and often suppress their own reproduction. Kin selection is clearly an important explanation for such behaviour, but a possible alternative is group augmentation where individuals survive or reproduce better in large groups and where it therefore pays to recruit new members to the group. The evolutionary stability of group augmentation is currently disputed. We model evolutionarily stable helping strategies by following the dynamics of social groups with varying degrees of subordinate help. We also distinguish between passive augmentation, where a group member benefits from the mere presence of others, and active augmentation, where their presence as such is neutral or harmful, but where helping to recruit new group members may still be beneficial if they in turn actively provide help for the current reproductives ('delayed reciprocity'). The results show that group augmentation (either passive or active) can be evolutionarily stable and explain costly helping by non–reproductive subordinates, either alone or leading to elevated help levels when acting in concert with kin selection. Group augmentation can thus potentially explain the weak relationships between relatedness and helping behaviour that are observed in some cooperatively breeding species. In some cases, the superior mutualistic performance of cooperatively behaving groups can generate an incentive to stay and help which is strong enough to make ecological constraints unnecessary for explaining the stability of cooperatively breeding groups.