AI Financial Advisors Show Bias and Inconsistency, Study Finds

Generative AI platforms are delivering inconsistent and sometimes biased financial guidance, raising concerns about their growing role in personal finance decision-making, according to new research published in the Journal of Financial Planning.

Researchers from the University of Georgia and the University of Rome Tor Vergata tested seven widely used AI platforms—ChatGPT, Claude, Copilot, DeepSeek, Gemini, Meta AI, and Perplexity—using identical prompts about emergency savings, retirement withdrawals, and portfolio allocation. The only variable changed was the race and gender of the household described in each prompt.

Wide variation across platforms

The study, conducted by finance professors Swarn Chatterjee, Brenda J. Cude, and Gianni Nicolini, found emergency savings recommendations ranged from $19,500 to $37,500 across the seven platforms—a statistically significant spread for identical financial scenarios. Portfolio allocation guidance showed even greater divergence, particularly in recommendations for equity, cash, and alternative assets.

Retirement withdrawal advice proved more uniform, with most platforms suggesting the traditional 4 percent rule. However, two platforms recommended a 5 percent withdrawal rate for certain demographic profiles while advising 4 percent for others with identical financial circumstances.

Demographic bias in some tools

Several platforms produced identical recommendations regardless of the household's demographic profile. Others did not. DeepSeek recommended substantially higher equity, cash, and alternative asset allocations for White male and White female households compared to African American male households. For African American male households, the platform suggested a 75 percent bond allocation versus 30 percent for the other groups.

Meta AI assigned the highest equity allocation to African American male households while recommending the highest bond allocation for White female households—a pattern that lacked clear financial justification.

Why it matters

As AI tools increasingly handle personal finance queries and some platforms now allow direct bank account connections, these inconsistencies carry real-world consequences. The researchers emphasized that generative AI systems "do not truly think, reason, or understand" the information they produce, and minor phrasing changes can significantly alter outputs. Unlike human financial advisors, these tools operate without fiduciary obligations, meaning users have no legal guarantee the guidance serves their best interests.

U.S. banking regulators have begun scrutinizing how financial institutions deploy AI in higher-risk functions. The findings suggest consumers should view AI-generated financial advice as a starting point rather than a replacement for credentialed professionals.

All platforms were tested through their free tiers during the same week in August 2025, according to Quartz, which first reported the study's findings.