Consumers are increasingly turning to artificial intelligence for financial guidance, and new research suggests the advice they receive depends heavily on which chatbot they ask, and in some cases, who they say they are.
A study published in the Journal of Financial Planning tested seven widely used generative AI tools, including ChatGPT, Claude, Gemini, Copilot, DeepSeek, Meta AI, and Perplexity, by feeding each identical household scenarios and asking for recommendations on emergency savings, retirement withdrawal rates, and investment portfolios. The researchers, from the University of Rome Tor Vergata and the University of Georgia, then varied only the race and gender of the household head to see whether the recommendations changed.
The results should get the attention of anyone in the accounts receivable management industry whose consumers may be using these tools to decide how much to save, spend, or pay toward their debts.
The gap between tools was substantial. Emergency savings recommendations for the same family ranged from $19,500 to $37,500, a statistically significant spread. Portfolio allocation advice varied just as widely, with recommended equity stakes ranging from 15% to 45% for an identical household. Only retirement withdrawal guidance was relatively uniform, with most tools defaulting to the traditional 4% rule.
More troubling were the demographic findings. Several tools produced different recommendations when only the race or gender of the household head changed. DeepSeek recommended a dramatically more conservative portfolio, with 75 percent in bonds, for an African American male-led household than for otherwise identical White male or female-led households. Other tools showed smaller but measurable variations in emergency savings figures across demographic groups.
The bigger picture: A 2025 survey cited in the study found more than half of U.S. adults with household incomes between $50,000 and $200,000 name the internet as their top source of financial guidance. As those searches shift to AI chatbots, consumers in collections may increasingly arrive at conversations with agents carrying AI-generated assumptions about what they can afford to pay, how much they should hold in reserve, and what a reasonable repayment plan looks like.
