麻省理工学院斯隆管理学院研究团队对大语言模型提供的金融建议进行了深入评估[1]。研究人员模拟1000名成年人使用GPT-5.2、GPT-5.6或Gemini 3 Flash获取金融建议，发现AI建议总体上鼓励储蓄、投资多元化和风险管理，但在应对失业等突发事件时表现不佳[1]。按照AI建议，30岁以上人群可获得可观的储蓄缓冲[1]。
然而，AI建议的质量存在显著的个体差异[1]。女性和金融素养较低的用户按建议行动后，到60岁时财富会减少约5万美元（占比4%）；缺乏AI使用经验的用户财富减少则接近10万美元（占比6%）[1]。结构化学术提示（包含年龄、工作状态、收入等完整信息）能显著改进AI建议质量[1]。
此外，研究还发现AI在推荐具体金融产品时存在偏差[1]。Vanguard产品在6%的大语言模型回复中被推荐，iShares产品在3.4%中被推荐，而这两家公司在提示词中的提及率都不足0.4%[1]。该项研究成果以《AI Financial Advice: Supply, Demand, and Life Cycle Implications》为题发表，并获得2026年瑞士金融研究所杰出论文奖[1]。
Researchers at MIT's Sloan School of Management have evaluated the quality of financial guidance provided by large language models, simulating scenarios involving 1,000 adults using GPT-5.2, GPT-5.6, or Gemini 3 Flash [1]. The study reveals that AI-generated advice generally encourages prudent financial behaviors including savings accumulation, portfolio diversification, and risk management [1]. However, the research identified significant disparities in outcomes based on user characteristics and input quality, with consequences that can reach approximately $50,000 in retirement wealth differences [1].
The analysis found that women and users with lower financial literacy experienced wealth reductions of about $50,000 (4%) by age 60 when following AI recommendations [1]. Users lacking experience with AI tools fared even worse, seeing their wealth decline by nearly $100,000 (6%) at retirement age when acting on the advice [1]. Meanwhile, adults over 30 who heeded AI recommendations achieved substantial savings buffers [1]. Critically, the AI systems performed poorly in scenarios involving sudden employment disruptions, suggesting limitations in handling unexpected financial emergencies [1].
The quality of financial advice improved significantly when users provided structured, comprehensive information including age, employment status, and income [1]. Additionally, the research uncovered instances of product bias: Vanguard products were recommended in 6% of LLM responses and iShares products in 3.4%, despite fewer than 0.4% of user prompts mentioning these companies [1]. The research, published as "AI Financial Advice: Supply, Demand, and Life Cycle Implications," earned the 2026 Swiss Finance Institute Distinguished Paper Award [1].