Half of Americans Now Ask AI for Financial Advice, but How Good Is It?

New MIT Sloan School of Management research finds that while AI can improve saving and spending guidance over a lifetime, it still reflects biases tied to user inputs and struggles with portfolio rebalancing.

Millions of households have turned to generative AI instead of human financial advisors for financial guidance, with more than half of adults in the United States and the United Kingdom having asked for advice, likely more than the share who consult a human financial advisor. The technology could make financial guidance far more accessible for people who can’t afford professional advice. But how good is it?

In a new working paper, “AI Financial Advice: Supply, Demand, and Life Cycle Implications,” MIT Sloan School of Management assistant professor Taha Choukhmane and his co-authors developed a novel way to measure and analyze the quality of AI’s financial guidance, finding that AI generally gives people sensible advice across the course of their lives.

“Generative AI has real potential to improve access to financial advice and support better decision-making over a lifetime,” said Choukhmane. “At the same time, its impact is not uniform.”

Read More on Fintech : Global Fintech Interview with Rob Young, Managing Director – UK at InDebted

How does AI’s financial advice stack up?

Choukhmane, along with Matthew Akuzawa and Weidong Lin from MIT, and Tim de Silva from Stanford University, built a simulation of how people earn, change jobs, invest, and pay taxes over their lives. They asked 1,000 participants to write three prompts to an LLM financial advisor: describe their situation, describe their situation, ask how much to save versus spend, and ask how to invest. About half of the participants had recently used AI for financial advice or information.

The team created “virtual people” with different incomes, savings, and job statuses, and then simulated lifetimes from age 22 to 90, repeatedly feeding the human-written prompts from people in similar situations into ChatGPT 5.2 and Gemini 3 Flash, and translating their responses into financial decisions. Simulations included random life events or “shocks.”

What were the results of using AI for financial advice?

Choukhmane and colleagues found that, compared with real-world behavior, AI pushed people closer to standard economic models of how people should make financial decisions over their lives. It consistently recommended smoothing spending over time by saving more during working years and drawing down in retirement, and investing heavily in diversified stock funds but reducing exposure after age 45.

Under AI guidance, most of the simulated people over age 30 built meaningful savings buffers, diversified into stock funds, and shifted from equities toward safer assets as they aged. In many cases, simulated wealth exceeded $1 million by retirement.

“It’s actually surprising the advice aligns with life-cycle theory, since these systems aren’t designed to optimize financial decisions, and they could easily end up reinforcing bad habits or telling people what they want to hear,” Choukhmane said. “Taken together, the results suggest AI could offer low-cost, widely accessible financial guidance that helps offset some limitations of traditional advisers, including high fees, biases, and conflicts of interest.”

AI also introduced topics that users didn’t raise. Liquidity appeared in 83% of the AI’s responses, despite only 6% of people mentioning it. Saving showed a similar gap: only 20% of people mentioned it in their questions, while AI included it in 76% of its answers. AI also frequently suggested safer, more diversified investment options such as high-yield savings accounts and government bonds.

“AI does more than just echo user prompts, it expands the conversation, introducing additional options and, at times, even identifying specific investment products that were not part of people’s original query,” said Choukhmane.

What are the limitations of using AI for financial advice?

AI struggles to adjust spending after income shocks; it passively shifts portfolios rather than actively rebalancing them; and it recommends too little gradual drawdown in retirement. Outcomes also varied sharply across groups. Following advice generated in response to prompts written by people with low financial literacy produced nearly $50,000 (4.11%) less wealth at age 60 than advice from prompts by high-literacy users, holding income and asset returns constant. The gap was even larger for those with prior AI experience: advice from prompts by people without any experience seeking financial guidance from AI generated nearly $100,000 (5.71%) lower wealth.

When it comes to gender, advice from prompts written by women led to nearly $60,000 (4.10%) less wealth than advice from prompts by men, largely driven by lower recommended equity exposure and less active rebalancing. The researchers decomposed this gap into two channels. About two-thirds came from differences in how men and women write their prompts, even when researchers told the AI both prompts came from the same gender. For example, women tended to use words like “family,” “grocery,” and “pay” more often, while men leaned toward “strategy,” “crypto,” and “growth.”

The remaining third came from the AI treating gender itself as a signal, recommending more stock exposure when the same prompt was labeled as coming from a man rather than a woman. When the researchers inserted “I am a man” or “I am a woman” into otherwise identical prompts, the model recommended less equity to women. These differences could reflect reasonable guesses from the AI about how preferences or circumstances vary by gender (which the AI could ideally make explicit to users), the researchers note, or biased patterns learned from training data.

What can people do to get better financial advice from AI?

The findings suggest AI doesn’t substitute for financial knowledge. By testing the impact of different prompts, the researchers found that prompting the AIs with richer instructions   drawing on life-cycle planning, modern portfolio theory, and assumptions about a person’s finances and the broader economy   improved spending and saving advice and reduced reliance on the simple rule-of-thumb approach. However, it did not ultimately lead to better portfolio rebalancing recommendations.

Financial literacy is a key input to writing better prompts, so an effective use of AI could be to build financial understanding rather than passively accept its advice. The research also offers a reusable framework for evaluating AI financial advice, along with simple real-world benchmarks, like smoothing spending, spreading risk, and rebalancing portfolios, to keep future tools honest and improving.

“While these results reflect today’s models, the bigger picture is likely to persist: what people get from AI depends as much on how they ask as on how the system responds,” Choukhmane said. “That means access to good financial outcomes may depend not just on better models, but on better questions, and who is asking them.”

Catch more Fintech Insights : Finance as a Feature: The Monetization Shift in Global FinTech Platforms

[To share your insights with us, please write to psen@itechseries.com ]

The post Half of Americans Now Ask AI for Financial Advice, but How Good Is It? appeared first on GlobalFinTechSeries.