Understanding the socio-political attitudes of other people is a crucial skill, yet the neural mechanisms supporting this capacity remain understudied. This study used multivariate pattern analysis to examine patterns of activity in the default mode network (DMN) while participants assessed their own attitudes and the attitudes of other people. Classification analyses indicated that common patterns in DMN regions encode both own and others' support across a variety of contemporary socio-political issues. Moreover, cross-classification analyses demonstrated that a common coding of attitudes is implemented at a neural level. This shared informational content was associated with a greater perceived overlap between own attitude positions and those of others (i.e.attitudinal projection), such that higher cross-classification accuracy corresponded with greater attitudinal projection. This study thus identifies a possible neural basis for egocentric biases in the social perception of individual and group attitudes and provides additional evidence for self/other overlap in mentalizing.