Often included within 'high-risk sibling' studies, the Autism Observation Scale for Infants (AOSI) has only one independent replication study and no evaluation with community-ascertained cohorts. We administered the AOSI and established clinical measures with 103 infants (68% male) at 'high autism likelihood' on the Social Attention and Communication Surveillance - Revised (SACS-R) tool, at 9-14 months of age and again 6months later. AOSI Total scores showed adequate internal consistency and strong inter-rater agreement (live- or video-coded) and were approximately normally distributed at each visit. Modest significant associations presented between Time 1 AOSI scores and concurrent developmental/adaptive skills measures. Concurrent associations were stronger at Time 2, particularly between AOSI Total and Autism Diagnostic Observation Schedule (ADOS) Social Affect scores. AOSI scores were only moderately associated across Time 1 and 2 assessments, as were Time 1 AOSI with Time 2 ADOS scores. These data from a clinically indicated cohort broadly replicate previous AOSI validity accounts from 'high-risk sibling' studies, particularly beyond the first year. Strong inter-rater agreement indicates viable AOSI inclusion within protocols necessitating blinded evaluation (e.g. intervention trials). Moderate within-participant stability suggests that, like 'high-risk siblings,' community-ascertained infants experience variable early trajectories. [This article was co-written by the AICES Team.]