Abstract
This study examines how matching Embodied Conversational Agents' (ECAs) appearance to message emotional tone enhances eHealth persuasion through the lens of Elaboration Likelihood Model (ELM) and Social Cues Theory. Using Event-Related Potential (ERP) measurements with 42 participants, we found professional ECAs delivering neutral messages elicited neural signatures of reduced cognitive conflict (smaller N400) and increased attention (larger LPP), demonstrating central route processing through credible source cues. Conversely, positive messages paired with casual appearances leveraged peripheral route persuasion via social rapport. Behavioral data confirmed these patterns, showing highest persuasion when professional appearance aligned with neutral tone - a congruence effect explained by both theories. Results provide actionable insights for designing persuasive ECAs in healthcare contexts by strategically combining visual and textual cues to optimize either credibility or approachability based on communication goals. The integration of neural and behavioral measures offers novel evidence for how multimodal cue matching operates in digital health interventions.