AI Gives Better Financial Advice Than Expected, Researchers Say: There's One Big Catch

It is a quiet revolution happening behind millions of glowing screens. Across the UK and the US, ordinary people are no longer turning solely to traditional bank managers or expensive wealth advisors. Instead, they are asking chatbots how to manage their life savings.

Half of Americans now use artificial intelligence for financial guidance. Yet, until recently, nobody truly understood whether these digital assistants were building fortunes or risking futures.

A landmark study by Taha Choukhmane, Weidong Lin, and Matthew Akuzawa from the MIT Sloan School of Management, alongside Tim de Silva from the Stanford Graduate School of Business, has delivered a startling answer. The algorithms are remarkably good at managing money. However, they carry a single, critical flaw that could leave vulnerable savers exposed.

A Digital Lifeline for Working Families

Traditional financial advice is expensive. For decades, lower-income households have been priced out of professional wealth management. This dynamic leaves millions without basic strategies for retirement.

The research reveals that large language models act as a powerful equaliser. When researchers simulated decades of financial choices using models like GPT-5.6 and Gemini 3 Flash, the results caught scholars off guard.

The algorithms consistently urged users to build cash buffers, diversify their stock holdings, and reduce risk as they aged. Choukhmane, an assistant professor of finance at MIT Sloan, admitted the research team was surprised by the quality of the guidance.

For the average worker above age 30, following AI recommendations generated substantial safety nets. The systems pushed users toward higher savings rates during their working years and advised a sensible drawdown in retirement. In short, the technology delivered sound, fundamental economics to people who might otherwise have had no advice at all.

The Fatal Flaw in Algorithmic Logic

Despite these successes, the study uncovered a dangerous limitation. AI struggles when real life gets messy.

Human lives rarely follow a smooth curve. People face sudden layoffs, medical emergencies, and economic shocks. When the researchers introduced simulated job losses, the chatbots failed to adapt properly.

Instead of drawing sensibly from existing savings, the AI advised users to slash their spending too aggressively. The systems relied on rigid rules of thumb rather than nuanced judgment. Furthermore, the algorithms allowed investment portfolios to drift over time, failing to trigger active rebalancing when market conditions changed.

Without human oversight, an individual relying entirely on digital counsel could face severe financial strain during a personal crisis.

The Prompt Gap and Hidden Bias

The quality of AI advice depends heavily on who is asking the questions. The study revealed that how a user frames a prompt creates massive differences in final wealth. When researchers tested structured academic prompts, which included full details on taxes, income, and living costs, the AI performed exceptionally well. However, ordinary consumers do not write like finance professors. A typical user simply asks where to invest a spare $50 a month.

This discrepancy opens a troubling wealth gap:

• The Gender Gap: Men tended to use terms like strategy and growth, while women frequently mentioned family and pay. The AI recommended higher stock allocations to men, compounding into roughly $50,000 less wealth for women by age 60.
• The Literacy Divide: Users with prior AI experience or higher financial literacy received better advice, accumulating up to 6% more wealth near retirement than first-time users.

Part of this gap stems from user phrasing. Yet, Choukhmane noted that the model also altered its advice simply when a prompt was labelled as coming from a woman. Whether this reflects statistical probabilities or ingrained training bias, the outcome remains troubling.

Reimagining the Future of Wealth Management

The findings signal a massive shift for financial institutions. The study noted that chatbots frequently recommended specific investment products, such as Vanguard or iShares funds, even when users never mentioned those brands. Traditional marketing may soon matter less than how an algorithm perceives a financial product.

For consumers, the takeaway is clear. AI is a remarkably cheap tool to build basic financial understanding. However, it cannot yet replace human nuance during a crisis. The smartest approach is hybrid. Use the algorithm to construct daily discipline, but keep a qualified human in the loop when life throws a curveball.