The Humanness Premium: How Anthropomorphism and Information Accuracy Shape Trust, Credibility and Visual Attention Towards an AI-Generated News Anchor

AI-generated news anchors can now look and sound human. It is unclear whether a human-like anchor makes viewers trust what it says, even when the information is wrong. In a 3 × 3 within-subjects design, 92 adults watched nine 8-s news clips in which a synthetic anchor (low, medium or high anthropomorphism) delivered a claim about physical activity and mental health that was accurate, hedged or overclaimed. Trust, credibility and perceived anthropomorphism were rated after each clip, and gaze was recorded with webcam eye tracking. Trust and credibility rose from the low to the medium to the high anthropomorphism level, with diminishing returns. Accuracy had smaller effects: the overclaimed claim was rated least credible, whereas the accurate claim was not rated above the hedged claim. The anthropomorphism effects were accounted for by perceived anthropomorphism. Viewers looked longer at the newsroom video wall when the anchor was highly human-like (Holm p = .039), whereas dwell on the anchor did not differ after correction. Within participants, attention to the anchor was associated with higher credibility. The overclaimed claim from the high-anthropomorphism anchor was nonetheless rated above the accurate claim from the low-anthropomorphism anchor. Human-like AI anchors therefore need accuracy safeguards.