Earlier studies demonstrate reduced illusion strength in the Shepard illusion in adults and adolescents with an autism spectrum disorder (ASD) and in typically developing (TD) adults with high levels of autistic traits. We measured the strength of the Shepard illusion in ASD and TD children and tested if ten different eye-tracking measurements could predict group differences in illusion strength. The ASD children demonstrated reduced illusion strength relative to the TD group. Despite this, there were no mean differences on any of the eye-tracking measurements between groups. Even though none of the eye-tracking measurements revealed mean differences between the two groups, the degree to which spatial attention was directed toward the standard stimulus, as indexed by the number of saccades within and toward this stimulus, predicted the strength of the illusion in the overall sample. Furthermore, this active scanning of the standard stimulus was found to enhance illusion strength more strongly in the ASD than the TD group. Together, we conclude that scan patterns and the degree to which participants are able to shift between different locations in a visual scene did not account for group differences in illusion strength. Thus, the reduced strength of the Shepard illusion in ASD does not appear to be driven by how attention shifts or is spatially allocated. Rather, differences may relate instead to perceptual mechanisms that integrate visual information. Strategies that may aid ASD individuals to see this illusion more strongly could have them make even more eye movements within and between the stimuli presented in the illusion display.