ChatGPT and stock returns

Abstract This paper investigates the dynamic relationship between investor attention to generative artificial intelligence, proxied by Google search trends for ChatGPT, and U.S. equity market dynamics. Utilizing a structural vector autoregression framework with weekly data from December 2022 to February 2026, we analyze the conditional responses of equity returns and trading volumes across four major indices. We document a persistent predictive relationship, where positive shocks to information demand correspond to increased valuations for technology-oriented equities and declining valuations for traditional industrial sectors, indicative of a capital rotation effect. While the incorporation of aggregate stock volumes reveals a consistent decline in macro-level trading activity, supporting an immediate consensus repricing mechanism, supplementary firm-level estimations demonstrate significant volume spikes alongside targeted price movements. This confirms that the observed capital rotation is driven by highly specific, active capital reallocation toward direct AI integrators and away from perceived competitors and legacy physical capital, rather than merely reflecting index concentration. Extending the empirical framework to state-dependent specifications demonstrates that these stock market dynamics are highly conditional on the prevailing macroeconomic environment, including shifts in interest rates, economic activity, oil prices, and exchange rates. Finally, restrictive structural orderings reveal significant weakening in these predictive paths, confirming that information demand and asset pricing are simultaneously determined as interconnected dynamic relationships rather than strict, unidirectional causal effects.