Study Finds AI Platforms Give Inconsistent Financial Advice by Demographics

A study published in the June 2026 edition of the Journal of Financial Planning tested seven generative AI platforms on identical financial queries that differed only by the race or gender of the household head. Newsweek reported that the platforms were ChatGPT, Claude, Gemini, Copilot, DeepSeek, Meta AI, and Perplexity.

Researchers created standardized household profiles and asked each tool for guidance on emergency savings, retirement withdrawal rates, and investment portfolio allocations.

Most platforms recommended retirement withdrawal rates close to the 4 percent rule. Emergency fund recommendations ranged from approximately $21,000 to $37,500 for the same household information. Recommended stock allocations ranged from 15 percent to 45 percent.

DeepSeek produced some of the largest variations. It tended to recommend more conservative, lower-risk portfolios to women and some racial or ethnic minorities while suggesting relatively higher-risk, higher-return investments for otherwise identical investors described as men or members of majority groups.

Gemini, Claude, and Copilot provided the most reliable recommendations for investment portfolio composition, according to co-author Gianni Nicolini, professor of banking and finance at the University of Rome Tor Vergata.

"Our evidence is that, for each of these tasks, AI is useful to define a reference point that can be used to start a conversation with a professional financial adviser," Nicolini said. Co-author Brenda Cude, professor emeritus at the University of Georgia, noted that some platforms changed recommendations when only the demographic description changed.