Designing persuasive chatbots: Cross-market evidence on cognitive and affective pathways in AI-mediated retail interactions

Artificial intelligence enabled chatbots are increasingly embedded in retail customer journeys, yet evidence remains limited on how specific chatbot design features translate into customers’ willingness to recommend the retailer. Grounded in the Elaboration Likelihood Model, this study develops and tests a dual route persuasion framework in which chatbot cognitive cues, namely information depth and response quality, and peripheral cues, namely interface aesthetics, shape recommendation intention via two distinct user evaluations, perceived informativeness and perceived enjoyment, conditional on customer involvement. Two 2 × 2 scenario-based experiments were conducted in different retail contexts. Study 1 (N = 420) involved a Chinese online electronics retailer; Study 2 (N = 508) replicated the model with United States consumers interacting with an omnichannel fashion and home accessories retailer. Across both studies, cognitive cues increased recommendation intention indirectly through perceived informativeness, whereas peripheral cues operated primarily through perceived enjoyment; once these mediators were included, the direct effects of chatbot cues on recommendation were no longer significant, indicating full mediation. Involvement strengthened the cognitive cue to informativeness pathway and its associated indirect effect on recommendation, while leaving the enjoyment pathway largely unchanged. For retailers, the findings suggest two complementary design levers: prioritizing accurate, diagnostic responses, especially for highly involved customers, while also investing in interface features that sustain enjoyable interactions to stimulate positive word of mouth and recommendation.