Screening tools to improve identification of children living with HIV (CLHIV) have been validated and used in various settings. The aim of our study was to optimize a screening tool for Primary Healthcare Clinics (PHCs) in South Africa (SA). A cross-sectional study was conducted at PHCs in Johannesburg and Mopani Districts, between June 2021 and June 2022. Children 5-14 years of age with HIV negative or unknown status accompanied by their mothers, or appropriate caregivers, were enrolled. Demographic data, responses to the screening tool questions, and HIV test results were captured. Logistic regression modeling was used to optimize an existing 10-item screening tool, and sensitivity, specificity, and number needed to test (NNT) used to choose the final tool. We enrolled 14,147 children in the study, with 62 children testing HIV positive (HIV positivity of 0.4%). The 10-item tool with a single positive response had a sensitivity of 91.9% and specificity of 43.3%. An optimal combination of 5-items with two positive responses had the lowest NNT of 72, 82.3% sensitivity and 74.2% specificity. Maternal HIV status alone, HIV positive or unknown, had a 95.2% sensitivity, 65.0% specificity and NNT of 84. The 1-item tool only would have missed 5% of CLHIV (N = 3) compared with the 5-item tool that missed 18% (n = 11). A 1-item screening tool asking about maternal HIV status can improve efficiency of testing of children in primary healthcare facilities in SA and improve identification of CLHIV who are not on treatment.