For cooperation to be beneficial, cooperators should be able to differentiate individuals who are willing to cooperate from free-riders. In the absence of kin or of familiar individuals, phenotypic similarity (e.g. in terms of language) can be used as a cue of how likely two or more individuals are to behave similarly (whether they will cooperate or free-ride). Thus, phenotypic similarity could affect cooperation. However, it is unclear whether humans respond to any type of phenotypic similarity or whether only salient phenotypic traits guide cooperation. We tested whether within-group, non-salient phenotypic similarity affects cooperation in 280, 3 to 10 year old children and in 76 young adults (mean 19.8 years old) in the UK. We experimentally manipulated the degree of phenotypic similarity in three computer-based experiments. We found no evidence of a preference for, or greater cooperation with, phenotypically similar individuals in children, even though children displayed ingroup preference. Conversely, young adults cooperated more with phenotypically similar than with phenotypically diverse individuals to themselves. Our results suggest that response to non-salient phenotypic similarity varies with age and that young adults may pay more attention to non-salient cues of diversity then children.