Background: Impaired emotion processing constitutes a key dimension of schizophrenia and a possible endophenotype of this illness. Empirical studies consistently report poorer emotion recognition performance in patients with schizophrenia as well as in individuals at enhanced risk of schizophrenia ("at risk"). fMRI studies also report consistent patterns of abnormal brain activation in response to emotional stimuli in patients, in particular decreased amygdala activation. In contrast, brain-level abnormalities in at-risk individuals are more elusive. We address this gap using an image-based meta-analysis of the fMRI literature. Methods: fMRI studies investigating brain responses to negative emotional stimuli and reporting a comparison between at-risk individuals and healthy controls were identified. Frequentist and Bayesian voxel-wise meta-analyses were performed separately, by implementing a random effect model with unthresholded group-level T-maps from individual studies as input. Results: Seventeen studies with a cumulative total of 677 at-risk individuals and 805 healthy controls were included. Frequentist analyses did not reveal significant differences between at-risk individuals and healthy controls. These results were replicated with Bayesian analyses, providing strong evidence for the absence of meaningful brain activation differences across the entire brain. Region of interest analyses specifically focusing on the amygdala confirmed the lack of group differences in this region. Conclusions: These results suggest that brain activation patterns in response to emotional stimuli are unlikely to constitute a reliable endophenotype of schizophrenia. We suggest that future studies rather focus on impaired functional connectivity as an alternative and promising endophenotype.