Study Finds AI Financial Advice Guides or Misleads Investors

A recent experimental study conducted at the University of Bayreuth demonstrates that artificial intelligence can significantly enhance or severely undermine investor decision-making, depending on the alignment of its recommendations with optimal financial outcomes. The research, led by Dr. Fabian Herweg and published in a CESifo working paper, analyzed the behavior of 3,700 participants who were asked to select between two fictional investment funds tracking identical stock market indices. Although both funds mirrored the same market performance, they featured divergent fee structures that created a complex trade-off, with one provider objectively offering the lowest total cost across all contribution levels. Participants made their selections either independently or with guidance from one of three advisory options: a neutral AI chatbot, a bank-affiliated AI chatbot, or a human financial adviser. The advisory inputs were experimentally controlled to deliver either correct or biased recommendations, and the study was fully incentivized to mirror real-world financial stakes. The results revealed a stark divide in AI efficacy. When unobstructed, approximately 64 percent of participants correctly identified the superior fund. The introduction of accurate AI guidance elevated this success rate to roughly 87 percent. However, when the same algorithmic tools promoted the inferior option, correct selection rates plummeted to 34 percent. The research further highlighted a pronounced vulnerability across all participant tiers, including individuals with higher financial literacy who typically demonstrated stronger baseline accuracy. Even among these experienced decision-makers, misleading AI recommendations reduced correct choices to below 50 percent. Notably, participants selecting smaller investment amounts, who generally possessed lower financial experience, saw their accuracy jump from 28 percent to over 74 percent when guided by correct AI advice. Compounding these risks, the study found that AI advice exerted a stronger persuasive influence than traditional human financial advisers, regardless of whether the guidance was beneficial or harmful. Attempts to mitigate bias through explicit transparency proved ineffective. When participants were informed that the bank-affiliated chatbot could generate financial profits from specific fund selections, their purchasing behavior remained largely unchanged. The data indicates that standard conflict-of-interest disclosures do not meaningfully counteract algorithmic persuasion. Researchers concluded that while high-quality AI advisory systems can deliver substantial value to investors lacking specialized financial knowledge, unregulated or misaligned algorithms pose a significant risk to market decision-making integrity. Simply disclosing AI involvement is insufficient to protect consumer interests. The study recommends that financial technology providers implement robust auditing mechanisms and algorithmic monitoring to ensure AI recommendations remain strictly aligned with user objectives rather than institutional profit motives. These findings underscore the necessity for regulatory frameworks that prioritize objective performance metrics and consumer-centric design in automated financial advisory platforms.