Talking Sustainability: How the Conversational Layer Shapes Recommendation Outcomes in LLM Chatbots

Outdoor tourism faces increasing challenges from overtourism, environmental degradation, and climate change, underscoring the need for effective visitor management and the promotion of sustainable choices. Digital technologies such as recommender systems and chatbots already implement sustainability features through eco-scoring and ranking. Still, research has largely emphasized these algorithmic approaches rather than sustainability framing within conversations. This paper investigates the conversational layer, the dialogue level at which chatbots present and frame recommendations, as a complementary mechanism for embedding sustainability in user interaction. We examine two key levers: system prompts that shape the chatbot’s role, style, and guiding values, and tool configurations that vary in functionality, from baseline search to sustainability-based re-ranking, and in description, from neutral wording to explicit nudging. We investigate whether variations in these levers affect the sustainability of recommendations and final choices. Our findings suggest that dialogue framing can effectively nudge users toward sustainable behavior, extending sustainability research in recommender systems from algorithms to conversations