Problems with joint attention (JA) are core features of Autism Spectrum Disorders (ASD). Here, we investigate how typically developing (TD) children and children with ASD respond to joint attention (RJA) and initiate joint attention (IJA) with a gaze contingent avatar. Thirty-one participants with ASD and 33 TD matched controls followed and directed the avatar's gaze to a series of referent images. Viewing times and recognition memory for the referent images were measured and compared between RJA and IJA conditions. Analysis of correctly identified target images suggests comparable target recognition between IJA and RJA conditions for both groups, but poorer overall recognition memory in the ASD group. Eye tracking data suggests different viewing strategies between the ASD and TD groups. These findings demonstrate the importance of considering processing time and saliency of referent objects when creating interactive social technology for children with ASD and further highlights the potential of interactive, gaze contingent social characters for enhancing our knowledge of joint attention.