Stanford-MIT researchers found AI chatbots give women 3 percentage points less equity exposure, costing them roughly $60,000 in simulated retirement wealth by age 60. Two-thirds of the wealth gap stems from women's prompt language, but one-third persists even when only the gender label changes in identical prompts. Users with low financial literacy ended up nearly $50,000 poorer by retirement, while AI newcomers trailed experienced users by nearly $100,000 in simulated wealth.
Don't wait: the analyst who called NVIDIA in 2010 just revealed his top 10 AI stocks. See the full list FREE now. A working paper circulating this summer quantifies a suspected problem: when women ask AI chatbots for investing help, they receive more cautious advice than men, with a compounding cost of real dollars.
The Stanford-MIT team behind AI Financial Advice: Supply, Demand, and Life Cycle Implications, released as MIT Sloan Working Paper 7377-26 in May 2026, simulates the life of a virtual saver following advice from GPT-5.2 and Gemini 3 Flash, and finds women end up with roughly $59,890 less simulated wealth by age 60. The precise dollar figure carries uncertainty (standard error of $31,185), but the direction is solid. Lead author Tim de Silva of Stanford Graduate School of Business, working with MIT Sloan's Taha Choukhmane, Weidong Lin, and Matthew Akuzawa, recruited 1,000 demographically representative U.S. adults through Prolific and asked each to write three prompts: describe their finances, ask for spending advice, and ask for investing advice.
Roughly half had recently used AI for financial guidance. The researchers then ran simulated careers featuring job loss, market volatility, and mortality risk. The underlying percentage effects are robust: women's simulated wealth at 60 comes in 4.10% lower in log terms (SE 1.61%), driven by a 2.94-percentage-point lower recommended equity share (SE 0.13pp).
A separate randomized-label experiment shows about two-thirds of the equity-share gap is demand-driven: women more often used words like "family," "grocery," "credit," and "loan," which steered the model toward liquid, safer assets. Men leaned on "portfolio," "equity," "strategy," and "crypto," which drew more aggressive recommendations. The remaining one-third persists even when prompts are identical and only the gender label changes.
The authors do not assign a cause: it could be training-data bias or the model pricing in women's longer life expectancy. The models overall nudged users toward habits most financial planners would applaud: stock-market participation, age-declining equity allocations, cash buffers. Women simply received a milder version of the same good advice.
Extract — continue reading at the source.
