Wired to believe: source-authority cues shape perceived AI capability and neural persuasion

AI advisors support judgement, yet users rely on credibility cues when evaluating AI-generated advice. Using fNIRS and behavioural measures, we examined associations of source-authority cues with perceived AI capability, judgement updating, and prefrontal processing. Seventy adults evaluated food-safety rumours before and after the same Pepper robot delivered corrections attributed to social media users, local news outlets, or national regulators. Corrections were followed by judgement change; descriptively, more scenario-level changes occurred in the corrective direction than in reverse. However, source condition did not significantly predict post-correction accuracy or successful correction. Prefrontal activation changed within conditions, but between-condition differences did not survive statistical correction. Exploratory functional-connectivity analyses revealed differentiated within-condition patterns. Orbitofrontal cortex activation correlated with perceived AI capability and trust, but not accuracy, suggesting an association with subjective advisor evaluation. These findings suggest that source cues influence perceived AI capability and are associated with aspects of prefrontal processing during AI-mediated correction.