Why AI Will Never Beat Passive Indexing — Precisely Because AI Exists

The idea that artificial intelligence would one day outsmart the stock market once felt inevitable. A machine capable of scanning decades of price history, parsing millions of filings, and learning every nuance of investor behavior seemed destined to outperform even the best human portfolio managers. Retail traders joined the excitement; surveys show that a meaningful share of them now ask chatbots for stock picks.¹ Yet the promise contains a trap. The more powerful and widespread AI becomes, the more it reshapes market behavior—often in ways that render its own predictions less reliable. The very tool meant to beat the benchmark may be the force that keeps the benchmark unbeatable.

The Paradox Behind AI Stock Picking

AI adoption in trading has surged. Patent filings for algorithmic strategies referencing AI jumped from less than one-fifth in 2017 to more than half of all such filings since 2020.² With so many firms using versions of the same tools, the edge from machine learning doesn’t last long. Once an anomaly is discovered—and exploited—it dissipates under the weight of competition. Quant managers refer to this as alpha decay, and it has only accelerated as computational power has grown.³

Machine-learning models, by design, learn from the past. They search price history, earnings patterns, sentiment shifts, and macro relationships to find hints about what might come next. Yet when enough traders rely on similar datasets and methods, their actions reshape the very landscape they’re trying to measure. The models begin training on a world that no longer exists.

Reflexivity and the Vanishing Edge

AI-driven funds also trade differently from traditional active strategies. AI-heavy ETFs tend to rotate their portfolios roughly once a month—dramatically more often than the typical active ETF, which turns over about once per year.² That constant churning reflects a frantic race for signals that evaporate as soon as they appear.

Markets now digest information with unprecedented speed. When the Federal Reserve releases minutes, stocks begin reacting within seconds.² By the 15-second mark, prices are already moving in the direction that historically took 15 minutes to unfold. AI isn’t just processing information—it’s accelerating the entire market’s reaction time, shrinking the window in which any strategy can hold an advantage.

The IMF notes that while AI can improve informational efficiency, it may also amplify volatility during periods of stress.² The 2020 pandemic selloff offered hints of that dynamic. AI-driven ETFs sharply increased their trading during the drawdown, exhibiting herd-like behavior that can turn small imbalances into large moves.²

This kind of feedback loop echoes George Soros’s idea of reflexivity—the notion that predictions shape market behavior, which then alters the original conditions being predicted. When dozens of AI models identify the same perceived opportunity, they crowd into it together. Their buying pushes prices higher, creating a self-fulfilling prophecy until reality snaps back. And by the time the model “learns” from that pattern, the edge is already gone.

As one Reuters analysis noted, generic AI models often “lean too hard on a pre-established narrative” and rely too heavily on past price action to forecast what comes next.⁴ In a market increasingly shaped by the algorithms trying to model it, historical patterns lose their predictive power.

Why Passive Indexing Keeps Winning

The recent era of extreme market concentration highlights another problem for AI stock picking. A small group of mega-cap companies—many tied to the AI boom—has driven a disproportionate share of market returns. In 2023, seven technology giants were responsible for the majority of the S&P 500’s advance.⁵ By 2025, AI-related firms accounted for an even larger share of market gains.⁶

An index fund owns all of these names automatically. Active managers, including AI-driven ones, often avoid overweighting expensive mega-caps because of valuation concerns or risk models.⁷ When leadership narrows to a handful of firms, that caution becomes costly. Historical data shows that active managers tend to underperform precisely during periods of narrow leadership.⁷

There is also the reinforcing effect of passive flows. Because index funds must match their benchmarks, they buy more of the largest stocks as their weights increase. That mechanical demand lifts prices, which further increases those stocks’ index weight, attracting still more passive flows.⁸ It becomes a momentum loop powered by indexing itself.