Generative AI tools produce inconsistent personal finance recommendations and, in some cases, deliver different guidance depending on the race or gender of the person described in a prompt, according to a new study.
The study, published last month in the Journal of Financial Planning, tested seven widely available GenAI platforms — ChatGPT, Claude, Copilot, DeepSeek, Gemini, Meta $META AI, and Perplexity — across three household financial scenarios: how much to keep in emergency savings, what withdrawal rate to use in retirement, and how to allocate an investment portfolio. Researchers then resubmitted the same prompts with only the race and gender of the household's lead changed.
The authors — finance professors Swarn Chatterjee and Brenda J. Cude of the University of Georgia and Gianni Nicolini of the University of Rome Tor Vergata — found that emergency savings recommendations ranged from $19,500 to $37,500 across the platforms, a spread that was statistically significant. Portfolio allocation guidance also varied widely, particularly in equity, cash, and alternative asset categories.
Retirement withdrawal recommendations were more uniform, with most tools aligning with the traditional 4 percent rule, though two platforms suggested a 5 percent withdrawal rate for some demographic profiles while recommending 4 percent for others.
On the question of demographic bias, some platforms produced identical recommendations regardless of whether the household was described as led by a White male, White female, or African American male. Others did not. DeepSeek, for instance, recommended substantially higher equity, cash, and alternative asset allocations for White male and White female households than for African American male households, while assigning the African American male household a bond allocation of 75 percent compared with 30 percent for the other two groups. Meta AI suggested the highest equity allocation for the African American male household and the highest bond allocation for the White female household.
The researchers noted that GenAI tools "do not truly think, reason, or understand" the information they produce, and that even minor differences in how a question is phrased can shift the output significantly. The study also raised concerns about the absence of any fiduciary obligation on the part of these tools, meaning users have no legal guarantee that the guidance serves their best interests. "GenAI-driven responses may sound confident but can still be incomplete, misleading, or incorrect," they wrote.
The findings are relevant as AI tools take on a larger role in personal finance. ChatGPT has launched personal finance tools that allow users to connect bank accounts directly and receive guidance, while U.S. banking regulators have begun scrutinizing how financial institutions deploy AI across higher-risk functions.
Ultimately, the researchers concluded that while GenAI can help consumers begin to navigate financial questions, it is not a substitute for advice from a credentialed professional. Each platform was tested through its free tier, with all prompts submitted within the same week in August 2025.