Ask two chatbots the same personal finance question using identical financial details, and you may get dramatically different advice. One might recommend an emergency fund that is nearly twice as large as what the other advises.
A study published in the Journal of Financial Planning in June 2026 found that recommendations varied significantly across AI platforms and, in several cases, also by the hypothetical user’s race or gender.
Researchers tested seven widely used generative AI platforms on identical personal finance prompts and found significant inconsistencies, demographic bias in certain recommendations, and a gap most users never consider.
Two out of three Americans who have used generative AI said they have tapped it for financial guidance, with that share climbing to 82% among both Millennials and Gen Z, CNBC reported.
Study exposes wide gaps in AI financial recommendations
Researchers Gianni Nicolini of the University of Rome Tor Vergata; Brenda Cude, a professor emerita at the University of Georgia (UGA); and Swarn Chatterjee, Bluerock professor of financial planning at UGA, tested free versions of ChatGPT, Claude, Copilot, DeepSeek, Gemini, Meta AI, and Perplexity.
All prompts were submitted in August 2025, covering emergency savings, retirement withdrawal rates, and investment portfolio allocation.
Emergency savings recommendations ranged from $19,500 to $37,500 across platforms, a spread the researchers found to be statistically significant.
Portfolio allocation guidance also diverged widely in equity, cash, and alternative asset categories, the study confirmed.
The platforms generally aligned with broad principles such as the conventional 4% retirement withdrawal guideline, the authors noted.
Yet the scale of variation in savings and allocation recommendations raised direct concerns about the reliability of chatbot-generated money advice.
Demographic bias surfaces in chatbot financial guidance
The researchers then resubmitted the same prompts with only the race and gender of the hypothetical individual changed.
ChatGPT, Copilot, and DeepSeek all recommended higher emergency savings for women and African American users than for their white male counterparts, the study found.
More AI:
- Goldman Sachs has blunt message for AI stock investors
- Microsoft CEO sends a blunt warning on AI and the tech ecosystem
- The next AI infrastructure race has nothing to do with chips
“If I’m a consumer, the recommendation I receive can vary simply based on which AI platform I’m using,” Chatterjee said in a University of Georgia press release.
Chatterjee explained that AI models combine vast datasets on human behavior and finance to make assumptions about individual users.
Those assumptions, such as expecting minority male users to face longer job searches, can lead the tools to recommend larger emergency funds.
The tools may prescribe larger financial cushions for certain demographic groups without explaining the reasoning behind those adjustments.
AI chatbots have no fiduciary obligation to users
A licensed financial advisor operating under a fiduciary standard faces legal consequences for putting personal interests ahead of a client’s needs, including regulatory penalties, civil liability, and criminal charges.
“What they don’t have is that fiduciary duty,” Andrew Lo, director of the Laboratory for Financial Engineering at the MIT Sloan School of Management, told CNBC. “They don’t have the ability to suffer consequences if they make a mistake to the same degree that a human advisor does.”
Lo also warned that large language models consistently generate responses that sound authoritative, regardless of whether the underlying information is accurate.
That risk is sharper when users press chatbots for personalized answers on topics such as taxes, retirement timing, or portfolio allocations, where accuracy depends on details the model may lack.
The Consumer Financial Protection Bureau’s 2023 report found that chatbots deployed by financial institutions for customer service sometimes provide inaccurate information. The CFPB warned that such errors can violate federal consumer financial protection laws.
Data privacy adds another layer of risk for AI money advice
Asking a chatbot for money guidance often requires sharing sensitive financial details, including income, account balances, and spending patterns.
That data can be stored indefinitely by the company providing the AI, and human reviewers may access those conversations, Empower noted in its analysis of chatbot privacy risks.
Certified financial planner Brenton Harrison told CNBC that relying on AI for financial guidance creates an inherent tension.
Looking to [AI] for advice implies you are giving it enough information to form an opinion and make a recommendation, and that’s a step further than I’d go with AI.
Research from cybersecurity firm Harmonic Security found that 4.37% of workplace AI prompts and 22% of file uploads to GenAI tools in the second quarter of 2025 contained sensitive company information.
Empower has advised users to leave out personal identifiers such as Social Security numbers, exact income, and account numbers when sharing information with AI chatbots and to frame questions in general terms.
What experts say about relying on AI outputs for money decisions
“Trust but verify,” Chatterjee said in the University of Georgia release. “AI gives people a starting point, not an ending point. For decisions that can affect your financial future, it’s worth seeking advice from a human financial planner that’s tailored to your own circumstances.”
A “Return on Advice” survey of 2,202 Americans in October 2025 found 76% say technology can provide financial information but not judgment or trust, Empower reported.
The Journal of Financial Planning study does not argue that AI has no place in personal finance.
Its authors concluded that generative AI holds potential to expand access to financial guidance. Still, they said, safeguards are essential, and users who treat chatbot outputs as authoritative advice are taking on risks they may not fully understand.
