The purpose of this study was to examine the utility of the Systematic Observation of Red Flags as an observational level-two screening measure to detect risk for autism spectrum disorder in toddlers when used with a video-recorded administration of the Communication and Symbolic Behavior Scales. Psychometric properties of the Systematic Observation of Red Flags were examined in a sample of 247 toddlers of 16- to 24 months old: 130 with autism spectrum disorder, 61 with developmental delays, and 56 typically developing. Individual items were examined for performance to create an algorithm with improved sensitivity and specificity, yielding a total Composite score and Domain scores for Social Communication and Restricted Repetitive Behaviors. Codes indicating clear symptom presence were collapsed to yield a count of the number of Red Flags for the overall scale and each symptom domain. Results indicated significant group differences with large effects for the Composite, both Domain scores, and Red Flags score, and good discrimination (area under the curve = 0.84–0.87) between autism spectrum disorder and nonspectrum groups for the Composite, Social Communication Domain, and Social Communication Red Flags score. The Systematic Observation of Red Flags provides an observational screening measure for 16- to 24-month-olds with good discrimination, sensitivity, and specificity. A cutoff of 20 on the Composite is recommended to optimally detect autism spectrum disorder risk.