55% of Americans trust AI with their money but should they?

Artificial intelligence has quietly moved from a novelty to a cornerstone of personal finance. Just a year ago, roughly 1 in 10 Americans said they used AI to help manage their money. Today, that number has jumped to 55%, according to a new survey from TD Bank making AI more widely consulted than traditional financial professionals. A 2025 Gallup poll found that fewer than 2 in 5 Americans actually turn to a licensed financial adviser for guidance.

That shift carries enormous weight. With more than half the country asking chatbots about savings, investments and retirement, trillions of dollars in financial decisions are now being shaped, at least in part, by Artificial intelligenceresponses. And while the technology has advanced rapidly, researchers are raising serious questions about how reliable those responses actually are.

What researchers found when they tested AI money advice

A team from MIT and Stanford University set out to find answers. They gathered 1,000 Americans and asked each person to write out a financial question they would realistically send to a chatbot. The prompts were then fed into AI systems, and researchers carefully analyzed the responses. Their findings are detailed in a working paper titled Artificial intelligenceFinancial Advice, Supply, Demand, and Life Cycle Implications.

On the surface, the advice was reasonable. The chatbots generally encouraged users to build emergency savings, invest in low cost index funds and reduce exposure to riskier assets as they aged all sound, standard recommendations.

But the research exposed meaningful gaps. The Artificial intelligence systems consistently struggled with more nuanced concepts. For example, rebalancing the practice of buying and selling assets to maintain a consistent level of risk in a portfolio was rarely addressed. And consumption smoothing, which involves balancing spending and saving across different life stages to protect one’s standard of living, was largely overlooked as well.

The quality of AI advice depends on who’s asking

Perhaps the most telling finding was this: the quality of Artificial intelligence financial advice is directly tied to the sophistication of the question being asked. Researchers divided the 1,000 prompts into two groups those written by people with lower financial literacy and those written by people with higher financial literacy.

The more informed the question, the more useful the answer. By modeling investment outcomes based on the Artificial intelligenceresponses, researchers calculated that a user who asked a more financially literate question could expect to earn roughly 5% more over time compared to someone who asked a less informed version of the same question. Researchers also found that people who were already accustomed to using Artificial intelligence for financial questions received meaningfully better advice than first time or occasional users.

Women may be getting steered toward less profitable advice

The research also uncovered a gender imbalance in Artificial intelligence financial guidance that researchers found difficult to fully explain. When chatbots identified a user as a woman, they consistently recommended more conservative investment strategies suggesting lower stock market exposure and safer asset allocations compared to what they told men.

Part of the difference came down to the questions themselves. Women were more likely to ask about debt management and lower-risk investments, while men more frequently asked about equities. But even when researchers posed the exact same question to Artificial intelligencesystems and simply changed the indicated gender of the user, the disparity persisted.

Researchers speculated that Artificial intelligence may be drawing on assumptions about women having more cautious financial objectives, or that the models may have absorbed gender biases from the data they were trained on. Whatever the cause, the pattern echoes a well documented problem in traditional finance, where women have historically received more conservative advice than men, often to their financial detriment.

AI-written financial articles are also falling short

A separate study from Harvard Business School and Boston College examined whether AI-generated financial content holds up when published at scale. Researchers analyzed thousands of articles posted on Seeking Alpha, a widely read investing platform known for its contributor driven content.

Using Artificial intelligence detection tools, they found that at one point, 13% of Seeking Alpha articles showed signs of Artificial intelligenceinvolvement. After the platform banned AI-generated content in 2023, that figure fell to 4%. The quality difference between human written and AI-written pieces was measurable: Artificial intelligence assisted articles drew fewer reader comments, were less likely to be selected as editor’s picks, and the stocks they covered generated lower trading volumes and weaker average returns.

Human written articles, researchers noted, tended to include personal experience, instinct and emotional context. AI-generated content, by contrast, was more impersonal and generic a distinction that appears to matter to readers and markets alike.

The broader warning is clear: as AI becomes a dominant voice in personal finance, the people most likely to benefit from it may already be the ones who need it least.