Alexul | Istock unreleased | Getty Images
When it comes to personal finance, the advice given by artificial intelligence can be inaccurate or demographically biased, and can vary widely depending on the specific program a consumer uses, according to a new academic research study.
The study examined seven “widely available” generative AI platforms and found “significant variation” in how GenAI answered prompts about emergency savings, asset allocation, and withdrawals from retirement portfolios.
Researchers looked at free access versions of ChatGPT, Claude, Copilot, DeepSeek, Gemini, Meta AI, and Perplexity.
“Although GenAI’s response may sound confident, it may still be incomplete, misleading, or inaccurate,” said a paper published last month in the Journal of Financial Planning and authored by finance professors at the University of Georgia and the University of Rome Tor Vergata in Italy.
Its “suboptimal” or biased output raises questions “about the consistency and fairness of GenAI-driven recommendations,” according to authors Swarn Chatterjee, Brenda Coode, and Gianni Nicolini.
The findings come as a majority of Americans rely on AI to manage their money.
According to an Intuit Credit Karma survey released in September, two in three Americans (66%) who have used GenAI say they have used it for financial advice. Gen Z and Millennials have a higher percentage, at 82% for each cohort.
Experts say AI is generally good at providing an overview of financial topics. For example, why it’s important to diversify your investments and why exchange traded funds are and aren’t better than mutual funds.
However, they said it has limitations and users should not blindly trust its output.
For one thing, experts say, the program can provide incorrect answers due to so-called “illusions” in the algorithm.
“One of my particular concerns about the LLM is that no matter what you ask, you’re always going to get an answer that sounds authoritative, even if it’s not,” Andrew Lo, director of the MIT Institute for Financial Engineering and principal investigator at the Computer Science and Artificial Intelligence Laboratory, told CNBC in a March interview.
“You have to be very careful when making very specific calculations of your own personal circumstances,” Lo said.
Additionally, the AI is sensitive to how users write prompts, so small differences in input can lead to variations in recommendations. Additionally, AI has no fiduciary responsibility to its users. This means that there is no legal requirement to provide financial advice that is in your best interests.
Other research studies have also pointed to the limitations of AI in personal finance.
For example, in one study in 2024, researchers investigated whether ChatGPT could provide financial advice.
They found that this could be a “first stop” for households seeking financial advice, but ultimately found that the recommendations were “general” and that specific relevant information was often overlooked.
“While we believe ChatGPT can serve as a starting point for providing or finding financial advice, its recommendations should be carefully scrutinized and evaluated,” according to a study published in the Journal of Risk and Financial Management.
A recent study published in the Journal of Financial Planning queried seven GenAI platforms in August 2025 using the same set of prompts.
The researchers presented the platform with three identical financial scenarios related to emergency savings, optimal withdrawal rates from retirement savings, and recommended investment portfolio composition.
Next, we used the same prompts, but changed the race and gender of the hypothetical individual to see if that would change GenAI’s recommendations.
They found “significant variation in guidance” across platforms when it comes to emergency savings and asset allocation.
“Although these tools often produced recommendations that were broadly consistent with common financial planning principles, such as the 4% retirement withdrawal rule, there were significant differences across platforms in suggested emergency savings and portfolio allocations,” the researchers wrote.
“The findings suggest that GenAl may serve as a useful starting point for consumers, but it complements rather than replaces professional financial advice,” they said.
Of course, the GenAI tool is “still evolving” and future studies may find different results, they said. Additionally, the output from a paid GenAI model may differ from the output from the evaluated free version.
