Recent research has shown that interactions between autistic people do not evidence the same social communication difficulties seen during interactions between autistic and non-autistic people. This raises questions about whether the social context (i.e. the respective autism status of the people interacting with one another) of an interaction affects observer’s ability to identify people as autistic. To examine whether the social context of an interaction also affects observers’ ability to identify autistic people in social interactions, we showed autistic and non-autistic raters videos (rater n=78; 39 autistic) and pictures (rater n=54; 27 autistic) of autistic and non-autistic people interacting (in own-neurotype and mixed pairs), and asked the raters to identify whether each member of the pair was autistic. Raters were able to identify autistic individuals in the stimuli at rates above chance. Dyad type, stimulus type (i.e. video or photo), and rater neurotype (as well as two- and three-way interactions between these factors) all affected identification accuracy, with raters displaying different patterns of accuracy across these factors. Our findings suggest that observers’ ability to identify autistic people depends on a number of socially contextual factors. There may also be an in-group guessing bias when attempting to identify someone’s diagnostic status.