1 hour ago
24
Ask a chatbot for financial advice and you might get a perfectly reasonable answer. But if you’re a woman, that “perfectly reasonable” answer could leave you roughly $60,000 poorer by age 60, according to new research from MIT.
A working paper from MIT Sloan School of Management found that AI chatbots systematically deliver more conservative financial guidance to women than to men, resulting in a simulated wealth gap of about 4.1% over a lifetime. The culprit is a cocktail of gendered language patterns and the AI’s own interpretive biases.
How the bias works
The study, co-authored by MIT Sloan assistant professor Taha Choukhmane alongside Stanford’s Tim de Silva and MIT researchers Weidong Lin and Matthew Akuzawa, ran simulations using models including ChatGPT 5.2 and Gemini 3 Flash. Prompts were sourced from roughly 1,000 participants, and the researchers tracked how the AI’s financial recommendations played out across an entire simulated lifetime.
When women asked for financial advice, the AI recommended lower equity allocations and less frequent portfolio rebalancing. Lower stock exposure over decades compounds into a significant shortfall, the kind of slow-drip disadvantage that’s easy to miss in any single interaction but devastating in aggregate.
Two-thirds of the gap came down to how men and women phrased their questions. Women in the study tended to frame financial queries around family and household concerns. Men were more likely to use language centered on investment strategies and growth potential. The AI, trained on oceans of human text, picked up on those patterns and responded accordingly, essentially matching tone with conservatism.
The remaining third of the disparity was even more troubling. When the AI detected explicit gender cues in a prompt, say a name or pronoun that signaled the user was female, it adjusted its recommendations toward more conservative allocations even when the underlying financial situation was identical.
More people trust chatbots than human advisors
The study notes that over half of adults in the US and UK have now sought financial advice from AI, a figure that actually surpasses those who consult human financial advisors.
The research paper, titled “AI Financial Advice: Supply, Demand, and Life Cycle Implications” and classified as NBER Working Paper 35574, did find that AI advice generally aligns with established economic life-cycle models. The chatbots correctly encouraged higher savings during working years, promoted equity diversification, and suggested lower-risk approaches as users neared retirement. On topics like liquidity management and portfolio diversification, the AI performed reasonably well.
But it stumbled in more nuanced areas. Optimal retirement withdrawal strategies and appropriate reactions to unexpected income changes, things like a sudden inheritance or job loss, proved difficult for the models to handle.
The prompt is the product
One of the study’s more practical takeaways is that the quality of AI financial advice depends heavily on how users ask for it. Women who used growth-oriented language in their prompts received advice much closer to what men received. That’s useful to know on an individual level, but it also highlights a systemic problem. The burden of correcting for AI bias falls on the people least likely to know the bias exists.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·