Does Search Heat Ignite AI‐Stock Volatility? Evidence From Korea's AI Supply Chain

Abstract

This study examines whether firm‐level search attention predicts returns or market turbulence in Korea's AI equities. Using daily Naver search and KRX data for 108 AI‐related and matched‐control firms from 2020 to 2026, the results show that search shocks coincide with abnormal trading activity and absolute price movements but do not reliably predict next‐day or 5‐day returns. Instead, they forecast realized volatility and high‐low trading ranges for up to 20 trading days, although their incremental forecasting value beyond HAR‐RV is modest. Common AI visibility raises average AI‐stock turbulence but does not significantly amplify the marginal search‐risk relation. After common‐attention and leader‐news controls eliminate the broad two‐leader cascade, a news‐orthogonalized Samsung search innovation still predicts broader attention among actual suppliers over the next three trading days. Overall, the findings suggest that search heat contains more information about near‐term turbulence and attention diffusion than about return alpha.