Why reject ChatGPT? Prompt strategy as keys to mitigate perceived creativity differences in tourism recommendation

This study aims to examine how recommendation agent type (human vs ChatGPT) influences tourists’ acceptance. Although cross-disciplinary research has identified algorithm aversion, existing studies have not adequately considered the enhanced agency of ChatGPT and lack explanatory mechanisms specific to destination recommendation contexts. Grounded in effort heuristic theory, this study tests whether perceived creativity mediates the relationship between recommendation agent and acceptance, and whether prompt strategy (zero-shot vs chain-of-thought) moderates the effects of recommendation agent on perceived creativity and acceptance. This study manipulated recommendation agent through scenario-based experiments, designing three experiments (two online and one on-site) to test all hypotheses and the conceptual model. A total of 626 valid questionnaires were collected. Results show that human recommendations elicit higher levels of perceived creativity and acceptance than ChatGPT recommendations. Perceived creativity mediates the relationship between recommendation agent and acceptance. Additionally, prompt strategy significantly moderates the effects of recommendation agent on perceived creativity and acceptance. Specifically, under chain-of-thought (vs zero-shot), the differences in perceived creativity and acceptance between the two agents become non-significant. This study establishes perceived creativity as a core mechanism explaining how recommendation agent type influences acceptance in destination recommendation contexts, extends the application of effort heuristic theory to AI-driven tourism recommendation scenarios and expands research on algorithm aversion via prompt strategy. Practically, this study provides actionable guidelines for enterprises, employees and society on promoting the adoption of ChatGPT in tourism recommendation.