People increasingly turn to artificial intelligence for financial advice, and half of Americans report using it, says Taha Choukhmane, MIT Sloan assistant professor and co‑author of a paper measuring LLM advice quality. Research shows following AI recommendations creates sizable savings buffers for individuals over 30.

AI advises saving, heavy stock investment, early retirement drawdown, and reduced equity exposure after 45. It struggles with shocks like unemployment and often lets portfolios drift, offering limited rebalancing. Structured prompts improve quality but still lack active rebalancing.

The study built a life‑cycle model of income, jobs, taxes and investments, then asked 1,000 adults to write prompts for GPT‑5.2, GPT‑5.6 or Gemini 3 Flash seeking spending and investing advice. Simulations covered ages 22‑89, comparing AI advice to real behavior and academic prompts. Results show LLMs provide affordable, unbiased guidance that outperforms expectations, though advice varies by user traits and can widen wealth gaps.

Advice differs by gender, literacy and AI experience, with men and literate users getting higher equity recommendations that yield roughly $50,000 more wealth by retirement, while women and less‑literate users fall about $50,000 behind. The authors say clearer, life‑cycle prompts improve advice and recommend using AI as a learning tool alongside human advisors.