From Yahoo Finance: 2025-06-07 08:30:00

An academic study found that ChatGPT had significant predictive power for economic outcomes in asset markets, with an average daily return of 0.38% and a compounded cumulative return of over 650% from October 2021 to December 2023. However, real-world frictions like transaction costs and fees could strain returns. About 76% of gains came from shorts, which can be more challenging due to fees and availability of shares.

Lopez-Lira’s experiment with ChatGPT was taken outside of academia by Autopilot, an investment app. He began providing the app with ChatGPT’s investment picks on a monthly basis, with Autopilot users linking their brokerage accounts to the selections. The process required providing ChatGPT with a wide range of information to make informed decisions based on macroeconomic and financial data.

AI models like ChatGPT, Grok, and DeepSeek were gradually given more freedom to make investment decisions, including deciding on weightings and selecting a broader range of positions outside initial parameters. The models were fed the same information to ensure consistency in decision-making. ChatGPT’s 15 picks were based on data until March 31, with weightings totaling 90% due to a 10% cash buffer left by the model.

Performance of the AI-managed portfolios varied, with ChatGPT’s portfolio returning 43.5% from September 2023 to May 30, 2025, outperforming the S&P 500’s total return of 34.7% over the same period. Grok’s portfolio returned 2.3% since inception on Feb. 11, while DeepSeek was down 0.25% since Feb. 3, both outperforming the S&P 500’s negative total return over the same periods.

While AI models show promise in mimicking professional portfolio management services, the lack of a major market event makes it difficult to determine their true effectiveness. AI models, while trained on historical data, may not be able to navigate unforeseen circumstances as humans can. Additionally, the AI-managed portfolios on Autopilot only rebalance monthly, limiting their ability to react to sudden changes in the market.



Read more at Yahoo Finance: A professor testing ChatGPT’s, DeepSeek’s and Grok’s stock-picking skills suggests stockbrokers should worry