Behavioral Responses to AI Financial Advisors: Trust Dynamics and Decision Quality Among Retail Investors

This study examines trust dynamics and decision quality among retail investors interacting with AI financial advisors. Using a convergent parallel mixed methods design incorporating surveys (n=428) and interviews (n=42), we investigate how trust forms, evolves, and influences investment outcomes. Results demonstrate that institutional affiliation (=0.68, p<0.001) and perceived competence (=0.57, p<0.001) significantly influence initial trust formation, while system explainability features substantially impact trust sustainability (high-explainability M=4.76 vs. low-explainability M=3.28, p<0.001, d=1.78). Longitudinal analysis reveals non-linear trust trajectories with distinct investor classes showing increasing (58.2%), stable (23.7%), or degrading (18.1%) patterns over nine months. AI-advised portfolios outperformed self-directed investments by 7.2% on risk-adjusted returns (Sharpe ratio: 0.83 vs. 0.65, p<0.01) with significant reductions in overconfidence bias (41.3%) and disposition effect (37.8%). Financial literacy moderates these benefits, with high-literacy investors showing smaller performance differentials between AI-advised and self-directed conditions compared to low-literacy participants. Transparency regarding system limitations demonstrates particularly strong effects on trust calibration, with 76.4% of participants in transparent conditions exhibiting appropriate reliance patterns versus 34.2% in non-disclosure conditions. These findings advance understanding of human-AI interaction in financial contexts and provide implications for designing trustworthy advisory systems that enhance retail investor outcomes.