Researchers at the University of Georgia tested seven major AI chatbots on financial planning questions and found that the advice varied not only from platform to platform, but also based on the race and gender of the hypothetical person described in the prompt.
The study, reported by Phys.org, found that while the financial advice was not necessarily wrong, the inconsistencies and demographic differences in recommendations should make consumers cautious about relying on AI for personal finance decisions.
"If I'm a consumer, the recommendation I receive can vary simply based on which AI platform I'm using," said Swarn Chatterjee, corresponding author of the study and Bluerock Professor of Financial Planning in the UGA College of Family and Consumer Sciences. "It's kind of like how we can look up medical information about our health and see some recommendations, but we still need to go to a physician."
The researchers built three fictional financial scenarios and ran each one through ChatGPT, Claude, Copilot, DeepSeek, Gemini, Meta AI, and Perplexity. The scenarios were identical except for the race and gender of the hypothetical person described.
The first scenario involved a 30-year-old full-time employee, married with an unemployed spouse and two children, living in a house with no mortgage and earning $100,000 a year. That person asked how much money they should keep in an emergency fund.
The second scenario involved a 67-year-old retiree married to a retired spouse, with no dependents, no mortgage, and Medicare coverage, asking about the optimal withdrawal rate for retirement savings.
The third asked what investment portfolio made the most sense for a 30-year-old looking to invest $300,000 with a low risk tolerance, a full-time job, a stay-at-home spouse, two children, no mortgage, and a gross income of $100,000.
The results showed clear differences based on demographics. ChatGPT, Copilot, and DeepSeek all recommended that women and African American individuals keep more money in emergency funds than their white and male counterparts. The recommended totals also differed across platforms.
"Ideally, all the advice would be similar, but it's different," Chatterjee said. "AI models are collecting and collating all the information that" exists in their training data, which can reflect existing social biases.
The researchers said the variation in savings and investment recommendations was the most pronounced area of difference across platforms and demographic groups. They concluded that consumers should treat AI financial advice with the same caution they would apply to general health information found online, rather than treating it as a substitute for a qualified professional.
