The human face communicates a wealth of socially relevant information such as person identity, emotion, and intention. A consistent behavioral finding in autism spectrum disorder (ASD) is reduced attention to or difficulty drawing meaning from faces. However, neuroimaging research into the neural correlates of face processing differences in ASD has returned mixed results. While many studies find hypo-activation or hypo-connectivity of core and extended face network regions, others show hyper-activation, equal activation, or even activation shifted to object-selective fusiform gyrus (FG) regions in ASD during face processing. This study consolidates two decades of literature to reveal common and consistent patterns of brain activation when viewing human faces in ASD. It also addresses whether face processing in ASD is routinely shifted to object-centric regions of the FG. To do so, we conducted an extensive search of the neuroimaging literature according to PRISMA guidelines. Peak activation coordinates from a final set of 23 studies, yielding a sample of 713 participants (338 ASD), were included for quantitative meta-analysis using Activation Likelihood Estimation (ALE). ASD within-group results across studies revealed a single activation cluster in the left FG, which presented laterally to the mid-fusiform sulcus (MFS). Typically developing groups displayed common activations across core and extended face network regions. Exploratory analysis of between group findings from the literature did not yield significant results. Overall, our results suggest that individuals with ASD consistently activate at least one typical face network region, the left FG, when processing faces and this activation is not routinely shifted to object-centric areas of the FG.