Skip to content
Book Open access

When Does AI Disclosure Build Trust? The Effects of Transparency Depth and Human Oversight on Reliance on AI Advice

Nov 2026 · 0 citations

Abstract

A basic AI label reveals that AI was involved but says little about how the resulting advice was produced or reviewed. This study tests a first-stage moderated mediation model in which transparency messages relate to self-reported willingness to follow AI-mediated financial advice through perceived trustworthiness, conditional on AI role. In a six-condition between-subjects experiment (N = 227), participants received the same portfolio recommendation across three transparency conditions (no disclosure, basic AI disclosure, and detailed explanation) and two AI-role framings (primary source and support tool). The no-disclosure conditions formed a common participant-facing baseline in which AI role was not communicated. Within the support-tool reference condition, basic disclosure did not increase perceived trustworthiness relative to no disclosure, whereas the detailed message increased trustworthiness relative to both no disclosure and basic disclosure. Perceived trustworthiness was strongly associated with willingness to follow the recommendation (b = 1.018, p <.001). The detailed-message × AI-role interaction was significant (b = −.291, 95% CI [ −.450, −.132]), and the detailed-versus-no-disclosure indirect effect through trustworthiness excluded zero in the support-tool condition (indirect effect =.504, 95% bootstrap CI [.032,.976]) but not in the primary-source condition. Because the detailed messages also differed in whether human review was present, the interaction captures differences between advisory-message configurations rather than a pure effect of explanation depth or AI role. The findings advance a content-sensitive account of AI transparency: labels alone were insufficient, while what transparency communicated about advice production, decision authority, and human review was consequential for trustworthiness judgments and willingness to follow AI-mediated advice.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.