Chatbot tone matters: upbeat for subjective tasks, neutral for factual ones, study finds

The researchers also analysed Weibo posts, which documented consumers responding differently to AI emotional expressions in subjective and objective service contexts. The authors say the results suggest firms should design AI service agents as social actors rather than purely functional tools.

The work covers only commercial service text and uses satisfaction as the single outcome, so effects on repurchase intentions or word-of-mouth remain untested.

The study tested how an AI agent's emotional expression, positive or neutral, interacts with the type of service task, subjective or objective, to shape consumer satisfaction. It began with Weibo posts, where consumers reacted differently to AI emotion depending on the service context, then moved to three scenario experiments.

The experiments showed a clear split. Positive emotion helped on subjective tasks and neutral emotion helped on objective tasks. Mediation analyses traced these effects to two separate routes: perceived experience for positive emotion and perceived agency for neutral emotion.

The authors conclude that firms can use emotional expressions to shape how consumers judge AI, but the design should align with the task. They describe the study as strong in internal validity but limited in external validity.