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

Key Findings

• — Privacy assurance increased perceived privacy protection (F(1, 764) = 159.30, p below 0.001, partial eta squared 0.172, d = 0.91).
• — Professional-boundary warning increased boundary awareness (F(1, 764) = 176.40, p below 0.001, partial eta squared 0.188, d = 0.96).
• — Boundary awareness was associated with lower overreliance risk and stronger professional help-seeking intention; calibrated trust was highest when both messages were present; structural model fit CFI 0.955, RMSEA 0.044.
• — The authors state the messages should not be read as prompts for greater chatbot use.

Methodology Notes

Vignette study (participants read chatbot scenarios rather than using a live chatbot), n = 768 college students in China; boundary awareness measured with five items on professional limits, non-substitution for qualified care and unsuitability for crisis use. Author affiliations are a judicial police college, three hospitals and a middle school rather than a psychology or HCI department; peer reviewed, CC BY; published 1 September 2026 (online date 10 August 2026).

Sources