AI Stock Picking: Can Machines Outperform Investors?

The rise of artificial intelligence (AI) has sparked discussions about its potential to revolutionize stock picking, raising the question: can machines outperform human investors? Recent analyses, including the 2025 Quantitative Analysis of Investor Behavior report by Dalbar, reveal that human investors often struggle to achieve optimal returns, making a compelling case for AI’s capabilities in this arena.

Dalbar’s report outlines the profound impact of investor behavior on mutual fund transactions. Since its inception in 1994, the organization has consistently highlighted that many investors would have fared better by sticking to low-cost index funds rather than engaging in the more complex game of buying and selling individual stocks. For example, the average equity fund investor earned 8.5 percent less than the S&P 500 in 2024, a trend attributed to psychological biases that lead to poor decision-making.

A notable study by Hendrik Bessembinder at Arizona State University analyzed nearly a century of stock market data and found that over half of publicly listed stocks had negative cumulative returns from 1926 to 2023. While a small fraction of stocks, like Altria Group, posted staggering returns—over 265 million percent—the median return was a stark negative 7.41 percent. These findings underscore the difficulty of stock selection, suggesting that many investors are overly optimistic about their abilities.

AI’s potential in stock picking has been put to the test through various experiments. Earlier this year, a project by Finder.com assessed ChatGPT‘s stock selection skills. The AI-generated portfolio achieved a gain of 4.9 percent, contrasting sharply with the average human-led fund, which lost 0.8 percent during the same period. Additionally, a high school student’s experiment using ChatGPT to invest in small-cap stocks yielded a remarkable 23 percent return, significantly outpacing the Russell 2000 index’s 3.9 percent increase.

Despite these encouraging results, the allure of stock picking remains deeply rooted in human psychology. Many investors are drawn to the thrill of trying to beat the market, often overlooking the challenges involved. The belief that one can successfully time the market or select winning stocks is a common fallacy. According to the Dalbar report, the “Guess Right Ratio,” which measures market timing success, stood at only 25 percent in 2024, tying for the lowest rate recorded.

AI strategies typically rely on established market patterns, which can lead to effective outcomes. A recent AI-driven experiment known as the “Comet Portfolio” constructed a conventional, tech-heavy investment strategy featuring major players like Amazon, Nvidia, and Microsoft. This approach aligns with momentum investing, which may help mitigate human biases such as fear and greed.

Nevertheless, the Dalbar report emphasizes that the crucial barrier for investors is not merely poor stock selection, but rather their psychological tendencies. Factors such as loss aversion, regret, and herd behavior often compromise returns. The report reveals a worrying trend: many investors tend to raise cash during market downturns, thus locking in losses.

The question remains whether AI can effectively counteract these emotional pitfalls. While AI systems promise to provide objective analysis, they are not immune to the biases inherent in the data and algorithms they are trained on. If the instructions fed into AI models reflect human biases, the technology may not outperform the market long-term.

Critically, even the most sophisticated AI may struggle to retain its position during market fluctuations. Investors have historically reacted to price corrections with panic, potentially “firing” an AI trader in times of distress. This reaction could undermine the potential advantages AI offers.

Despite its promising capabilities, AI has not yet demonstrated an ability to consistently outperform broad index strategies. Research indicates that timing-based investment strategies utilizing large language models tend to underperform passive benchmarks. AI may be overly cautious in bullish markets and excessively aggressive during bearish phases, leading to considerable losses.

In summary, while AI shows potential in stock selection and trading, its current track record is mixed, primarily reflecting established market trends rather than groundbreaking insights. Professional stock pickers and AI have yet to prove superior to straightforward index strategies. Investors should approach AI-driven investment advice with cautious optimism, recognizing that until consistent long-term results are demonstrated, skepticism remains prudent.

The insights from Dalbar’s 2025 QAIB study serve as a reminder of the importance of understanding the psychological factors that influence investment decisions. As the landscape of stock picking evolves, the dialogue between human investors and AI technology will undoubtedly continue.