Eye tracking (ET) holds potential for the early detection of autism spectrum disorder (ASD). To overcome the difficulties of working with young children, developing a short and informative paradigm is crucial for ET. We investigated the fixation times of 37 ASD and 37 typically developing (TD) children ages 4-6 watching a 10-second video of a female speaking. ASD children showed significant reductions in fixation time at six areas of interest. Furthermore, discriminant analysis revealed fixation times at the mouth and body could significantly discriminate ASD from TD with a classification accuracy of 85.1%, sensitivity of 86.5%, and specificity of 83.8%. Our study suggests that a short video clip may provide enough information to distinguish ASD from TD children.