Study finds humans still trust human financial advisors over AI

Researchers from Georgia Institute of Technology (Aryan Ramchandra Kapadia, Eshwar Chandrasekharan, Koustuv Saha) conducted a preregistered vignette experiment (N=285) to isolate how presentation style versus source attribution affects trust in financial advice. The team systematically varied communication styles between AI Financial Assistant, Certified Financial Planner (Expert), and Online Community Forum (OC) while keeping all substantive financial content identical—including facts, numbers, recommendation direction, and reasoning logic.

The study revealed a clear human preference: Expert advice was rated more favorably than AI advice on 9 of 10 outcomes (effect sizes d=0.20–0.47), even when source labels were removed. When AI advice was mislabeled as human-generated, its perceived situational fit and overall quality increased by d=0.42, while the Expert advantage in situational fit decreased by d=0.36. This demonstrates that attribution labels act as interpretive frames that shape trust more strongly than the underlying advice quality, suggesting disclosure policies must be carefully designed to avoid misleading perceptions rather than simply providing transparency.