Abstract Background and objectives A promising candidate for a vulnerability indicator for psychosis is the restricted scanpath. Restricted scanning of social stimuli, such as faces, might also contribute to misinterpretations of facial expressions and thus increase the likelihood of delusional interpretations. Moreover, similar to other vulnerability indicators of psychosis, scanpaths may be susceptible to stress. Thus, we hypothesized that scanpath restriction would increase as a function of delusion-proneness, stress and their interaction. Methods Participants were asked to look at neutral faces and rate their trustworthiness under a stress and a non-stress condition, while the eye gaze was recorded. The non-clinical sample was classified into low- and high-paranoia scorers using a median split. Eye-tracking parameters of interest were number of fixations, fixations within emotion-relevant facial areas, scanpath length and duration of fixations. Results In general, high-paranoia scorers had a significantly shorter scanpath compared to low-paranoia scorers ( F (1, 48) = 2.831, p  = 0.05, η p 2  = 0.056) and there was a trend towards a further decrease of scanpath length under stress in high-paranoia scorers relative to low-paranoia scorers (interaction effect: F (1, 48) = 2.638, p  = 0.056, η p 2  = 0.052). However, no effects were found for the other eye-tracking parameters. Moreover, trustworthiness ratings remained unaffected by group or condition. Limitations The participants of this study had only slight elevations of delusion-proneness, which might explain the absence of differences in trustworthiness ratings. Conclusions Restricted scanpaths appear to be partly present in individuals with subclinical levels of paranoia and appear to be susceptible to stress in this group. Nevertheless, further research in high-risk groups is necessary before drawing more definite conclusions.