Targeted screening of children at increased likelihood of autism is recommended. However, autism screening tools are usually validated for use mainly in low-likelihood populations. This study compared the accuracy of the Modified Checklist for Autism in Toddlers, Revised with Follow-up (M-CHAT-R/F), the ASDetect app, and the Social Attention and Communication Surveillance, Revised (SACS-R). Siblings of autistic children underwent autism screening at 12, 18 and 30 months old. At each visit, caregivers completed the M-CHAT-R/F and ASDetect while trained nurses tested the siblings using the SACS-R. At 36 to 48 months, the siblings underwent an Autism Diagnostic Observation Schedule--Second Edition (ADOS-2) assessment. 189 siblings were screened, 141 completed the study, and 32 were confirmed to have autism. Although not validated for use at 12 months, the M-CHAT-R/F had the best sensitivity among the three tools for this age group, suggesting that early signs are already apparent to caregivers. The M-CHAT-R/F had overall better sensitivity (0.72-0.83) across all age groups, but with overall lower specificity (0.55-0.77). The SACS-R and ASDetect had better positive predictive values at 18 and 30 months (0.60-0.68), while the M-CHAT-R/F was 0.43-0.48. Negative predictive values were generally high across all three tools across all age groups (0.78-0.93). Targeted screening of children at high likelihood of autism yielded a detection rate of 22.7% and should therefore be implemented routinely to facilitate early detection and intervention. The performance of autism screening tools should be examined in higher-likelihood populations for targeted screening of these children.