Scholars have argued that social evaluation, the capacity to evaluate different potential social partners, is an important capacity not just for humans but for all cooperative species. Recent work has explored whether domesticated dogs share a human-like ability to evaluate others based on prosocial and antisocial actions toward third parties. To date, this work has shown mixed results, suggesting that individual differences may play a role in dogs’ capacity to evaluate others. In the present study, we test whether attachment—an individual difference that affects human social evaluation performance—can explain the mixed pattern of social evaluation results observed in dogs. We first tested dogs on a social evaluation task in which an experimenter either helped or refused to help the dog’s owner open a container. We then assessed dogs’ attachment strength using a subset of the C-BARQ. We found that attachment was a statistically significant predictor of dogs’ preference toward the prosocial actor but was not a predictor in antisocial or control conditions. This finding provides early evidence that attachment may drive positivity biases in dogs and that attachment might explain mixed results within canine social evaluation literature.