Does humanlike AI nudge live-stream impulse buying? Dual moderation through consumers’ AI self-efficacy and AI self-competency

Purpose Drawing on the Heuristic-Systematic Model (HSM), the CASA paradigm and parasocial interaction (PSI) theory, this study examines how AI digital humans' anthropomorphic cues (empathy, social presence and personalization) drive impulsive buying in live-streaming commerce. It further investigates the dual-process moderating mechanisms of consumer AI literacy, namely, consumers’ AI self-efficacy and consumers’ AI self-competency. Design/methodology/approach Survey data from 527 Chinese consumers with recent AI-hosted live-shopping experiences were analyzed using partial least squares structural equation modeling (PLS-SEM). Findings Empathy, social presence (the strongest predictor) and personalization (a modest effect) significantly trigger impulsive buying. Moderation analysis further reveals a “rationality paradox”: consumers’ AI self-efficacy generates interactional fluency that paradoxically accelerates heuristic impulsive responses to these cues. Conversely, consumers’ AI self-competency triggers critical systematic processing, acting as a defensive buffer that mitigates impulsive tendencies. Originality/value This study advances the human-machine commerce literature by examining and explaining the rationality paradox. By integrating HSM, it extends traditional unidirectional PSI into bidirectional algorithmic synergy, demonstrating how distinct dimensions of AI literacy act as either propulsive accelerators or cognitive brakes in high-arousal environments.