When the Bot Stays Cold: How Non-Empathic Insurance Chatbots Shape Trust, Disclosure and User Experiences

This qualitative study examines how consumers experience a non-empathic chatbot during a simulated auto-insurance quote. Ten participants completed the interaction and then participated in a Retrospective Think-Aloud Protocol (RTAP) interview to articulate interpretations and emotional reactions at key moments. Interviews were analyzed using an inductive approach, and findings were triangulated with facial-expression data. Results converge around three aggregated dimensions (1) Trust, disclosure and reassurance in a non-empathic context, where requests for sensi-tive information were scrutinized through privacy risk, perceived opacity, and need for justification, and where the chatbot’s mechanical tone could reduce willingness to disclose; (2) Humanization and emotional disconnect, where the absence of ac-knowledgment after emotionally salient disclosures was often perceived as cold or missing, while anthropomorphic cues (photo/emojis) produced polarized reactions (helpful vs. suspicious/inauthentic); and (3) Efficiency vs. emotional resonance di-lemma, where participants prioritized speed and clarity, but still valued brief/task-linked cues and explanatory statements that improved comfort and perceived pro-fessionalism without slowing the process. These findings show how emotion-neutral chatbot design shapes trust, disclosure comfort/privacy perceptions and UX in a high-involvement/low-touch/data-intensive insurance service context. Meth-odologically, it shows how triangulating facial-expression data with qualitative feedback strengthens and nuances interpretive validity by revealing convergence and divergence across sources.