Purpose This study aims to identify the cues that influence artificial intelligence-generated content (AIGC) persuasion and to explore the dynamic mechanisms that shape persuasion during human–AI interactions. Design/methodology/approach Guided by the think-aloud method and grounded theory, this study used “ERNIE Bot 3.5” as the experimental tool to conduct think-aloud experiments and semi-structured interviews based on six search tasks. This process generated over 110,000 words of data from 23 participants. Three coders independently analysed the raw data, identifying 982 reference points, 59 initial codes, 21 initial categories and 7 main categories. Furthermore, a dynamic mechanism of AIGC persuasion was constructed by integrating the heuristic-system model. Findings This study reveals that both heuristic cues (technical features and interactive features) and systematic cues (content features) collectively influence AIGC persuasion through users’ interactive perceptions. Task types and information prediction were found to moderate this process. This study further suggests that while systematic cues play a dominant role in shaping AIGC persuasion, this dominance diminishes in more objective tasks, such as fact-based tasks, as the number of dialogue rounds increases. Originality/value From a dynamic perspective, this study reveals the cues influencing AIGC persuasion and proposes a theoretical model. These findings contribute to a deeper comprehension of the persuasion process in AIGC and offer valuable insights for optimising GAI models and user interaction design.
Sun et al. (Fri,) studied this question.