<p id=p00005>Prosocial behaviour, that is, behaviour intended to benefit others, has been linked to a variety of desirable traits, including positive relationships, better academic performance and lower levels of antisocial behaviours. As such, the origins and the mechanisms underlying the remarkable individual differences in prosocial behaviour are the focus of an increasing number of studies, with numerous research consistently documenting the important role of positive parenting and empathy. Notably, differentiating between cognitive and emotional components of empathy may help further clarify the processes by which parenting eventuates in prosocial behaviour. Although all children may be impacted by parenting, some children benefit more than others from good-quality rearing. Recent research has suggested that the oxytocin receptor (<italic>OXTR</italic>) gene rs53576 polymorphism could determine the degree to which a child is influenced by environment. The biological function of rs53576 polymorphism has yet to be delineated, and the literature is mixed with regard to heterozygote (AG) grouping; thus, the implications for AG grouping are not well understood. Therefore, the dummy coding, additive coding, dominant coding and recessive coding models were all investigated in this study to test the nature of gene effect. This study aimed to extend previous studies on the association between parenting and prosocial behaviour by examining the mediating role of cognitive and emotional empathy and the moderating role of the <italic>OXTR</italic> gene. <p id=p00010>The participants were 1082 mother-offspring dyads (adolescents’ mean age: 12.32 ± 0.48 years, 50.3% females) recruited from the community. At Time 1, mothers reported their positive parenting via the Chinese version of the Child-Rearing Practices Report (CRPR) and peer-rated adolescents’ prosocial behaviours. At Time 2, adolescent-reported perspective-taking and empathic concern, peer-rated prosocial behaviours and saliva samples were collected. All measures showed good reliability. Genotyping at <italic>OXTR</italic> gene was performed with MassARRAY RT software version 3.0.0.4 and analysed using the MassARRAY Typer software version 3.4 (Sequenom). <p id=p00015>Results showed that adolescents who received higher levels of maternal positive parenting exhibited more prosocial behaviours. However, the direct effect of positive parenting on prosocial behaviour became nonsignificant after controlling for baseline prosocial behaviour. Cognitive empathy, but not emotional empathy, mediated the association between positive parenting and prosocial behaviour. Specifically, positive parenting was positively associated with cognitive empathy, which in turn was positively associated with adolescent prosocial behaviour. Further, this mediation was moderated by the <italic>OXTR</italic> gene rs53576 polymorphism. For adolescents with AA and GG genotypes, positive parenting was related to higher levels of cognitive empathy, which increased prosocial behaviour. However, this mediation effect was not observed among adolescents with AG genotype. In addition, the results revealed evidence for an overdominance model for <italic>OXTR</italic> rs53576. Moreover, the G × E term predicted cognitive empathy but not prosocial behaviour. This finding suggests that cognitive empathy may be an endophenotype closer along the causal chain to the genotype and that the strength of the G × E effects was greater for empathy than for distal behavioural outcomes. <p id=p00020>These findings add to our understanding of how empathy and genetic factors contribute to adolescents’ prosocial behaviour within the family context. In addition, these results suggest that cognitive and emotional aspects of empathy are likely to be involved—in somewhat different psychosocial mechanisms—in the development of prosocial behaviour. Notably, the overdominance effect of <italic>OXTR</italic> should be interpreted with caution until replicated. However, when a three-category polymorphic genotype is used, as is commonly applied when modelling a dominant or recessive effect, both false positive and false negative results can occur, and the nature of the interaction can be misrepresented.