The prevalence of autism has increased over time, yet the rate of its detection remains static, often due to the scarcity of staff and resources necessary for supporting and diagnosing children with autism. Leveraging the advancing field of computing, this study introduces a machine-learning strategy to tackle this challenge. Unlike prior research which predominantly employed traditional machine-learning methods to diagnose older age groups, our focus was on identifying autism in toddlers, assessing the effectiveness of both deep learning and traditional machine-learning classifiers to achieve more accurate and efficient diagnostic outcomes. The dataset for this study, titled ‘Autistic Spectrum Disorder Screening for Toddlers,’ was compiled using the QCHAT-10 child assessment questionnaire. We examined five machine-learning models including Support Vector Machine, XGBoost, Logistic Regression, Decision Tree, and Multi-Layer Perceptron—a feedforward artificial neural network. These models were assessed using metrics such as accuracy, recall, precision, and F1 score, and underwent tenfold cross-validation to ascertain consistent and dependable mean accuracy scores. The comparative analysis identified the Multi-Layer Perceptron model as the most effective, with an accuracy of 98%. This research provides a potential auxiliary tool for psychologists and a self-assessment resource for individuals, aiding in the early identification of autism.