Aim: To investigate a novel observational rating protocol designed to expedite clinical diagnosis of autism spectrum disorder (ASD). Method: Two hundred and forty patients referred to a tertiary autism center (median age 8y 9mo, range 2y 6mo–34y 8mo; 188 males, 52 females) were rated using an adaptation of the Childhood Autism Rating Scale, Second Edition (CARS-2) based exclusively on patient observation (CARS-2obs). Scores were compared to expert diagnosis of ASD, parent-reported Social Responsiveness Scale, Second Edition (SRS-2) and, in a selected subset of patients, the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2). Results: CARS-2obs distinguished patients with a clinical diagnosis of ASD from those with non-ASD neuropsychiatric disorders (mean score=18 vs 11.7, p<0.001). Severity ratings on the CARS-2obs correlated with the ADOS-2 (r=0.68, ρ=0.64) and SRS-2 (r=0.31, ρ=0.32). A CARS-2obs cutoff point equal to or greater than 16 demonstrated 95.8% specificity and 62.3% sensitivity in discriminating individuals with ASD from individuals without ASD in a specialty referral setting. Interpretation: The CARS-2obs allows the rapid acquisition of quantitative ratings of autistic severity by direct observation. Coupled with parent/teacher-reported symptoms and developmental history, the measure may contribute to a low-cost diagnostic paradigm in clinical and public health settings, where positive results might help reduce delays in diagnosis, and negative results could prompt further specialty assessment. What this paper adds: The Childhood Autism Rating Scale, Second Edition based on patient observation distinguished individuals with versus without autism spectrum disorder (ASD). A score equal to or greater than 16 on this assessment showed high specificity for a diagnosis of ASD.