Privacy assurances and professional-boundary warnings in generative AI mental health chatbots: a randomized vignette experiment on calibrated trust, overreliance risk, and professional help-seeking intentions

Abstract

Introduction: Generative artificial intelligence (AI) chatbots are increasingly used by students to look up mental health information, seek reassurance, explore coping strategies, and identify possible sources of support. In such sensitive contexts, however, their use raises concerns about privacy, professional boundaries, and the risk of relying on AI beyond its proper role. This study examined whether two brief interface messages-privacy assurance and professional-boundary warning-affect students' safety-related evaluations of generative AI mental health chatbots. Methods: We conducted a 2 × 2 randomized vignette experiment with 768 college students. Participants were assigned to one of four chatbot scenarios that either included or omitted privacy assurance and professional-boundary warning. Measures covered perceived privacy protection, boundary awareness, calibrated trust, safe-use intention, overreliance risk, and professional help-seeking intention. Results: (302) = 742.65, CFI =0.955, TLI = 0.948, RMSEA = 0.044, and SRMR = 0.046. Perceived privacy protection and boundary awareness were associated with calibrated trust. Calibrated trust and perceived privacy protection were associated with safe-use intention, whereas boundary awareness was linked to lower overreliance risk and stronger professional help-seeking intention. Calibrated trust was highest when both messages were present. Discussion: These findings suggest that visible privacy and boundary messages can influence how students judge AI chatbots in mental health-related situations. Such messages should not be understood as prompts for greater use. Rather, they may help users treat chatbots as limited tools for information and support navigation and recognize when professional help is needed.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 1 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.