Can AI change how guests behave? Field experimental evidence on guilt appeals and sustainable choices in hotels

AI enabled service touchpoints are rapidly diffusing in hotels, yet causal evidence on whether they influence guests’ sustainable behaviour remains limited. We examine how robot concierge message framing affects in stay conservation in China. Study 1 uses 46 semi structured interviews to explore how guests interpret robot delivered sustainability prompts. Study 2 implements a cluster randomised field experiment over 12 weeks using difference in differences and hotel operational records of housekeeping opt out and towel and linen reuse. Guilt framed robot messages increase sustainable behaviour relative to neutral messages. Survey evidence identifies two mechanisms: anticipated guilt and perceived social presence. Effects are stronger among guests higher in face concern and environmental self identity. Sensitivity analyses support the findings while highlighting the inferential limitations of the eight hotel design. The findings show that AI moral prompts can function as socially interpreted service agents that generate measurable sustainability gains in hotels today.