<p>This study investigated the impact of generative chatbots integrated into pharmacy flipped classrooms on students’ higher-order thinking skills (HOTS: critical thinking, problem-solving, creativity) and explored the underlying mechanism. A pretest-posttest design tracked 213 pharmacy students over one semester. Data were collected via validated HOTS questionnaires (administered at pretest and posttest), a post-intervention survey on chatbot usage (metrics: frequency, duration, intensity; patterns: e.g., information retrieval, concept clarification), student interviews, and weekly quiz scores. Multivariate statistics revealed a significant improvement in HOTS from pre-test to post-test (<i>p</i> &lt; 0.001, medium effect size). This improvement was supported by two synergistic factors: well-structured teaching designs and students’ positive learning attitudes coupled with strong learning abilities. While quantitative chatbot usage metrics showed no significant association with HOTS gains (<i>p</i> &gt; 0.05), students’ perception of effective chatbot use was significantly associated with enhanced critical thinking (<i>β</i> = 0.789, <i>p</i> &lt; 0.001) and a moderate effect on problem-solving (<i>β</i> = 0.210, <i>p</i> = 0.008), but not with creativity, likely due to chatbots’ limited support for divergent thinking. Notably, qualitative data suggested that learning-oriented deep interaction with chatbots may serve as a mediating mechanism between effective use and HOTS improvement. These findings underscore that the educational value of chatbots derives from the quality, not the quantity, of their use. We recommend designing pharmacy-specific tasks to promote deep cognitive engagement and positioning chatbots as supplementary tools within the flipped classroom framework.</p>