The influence of generative artificial intelligence on financial market volatility, liquidity, and predictability

Purpose- This study examines the impact of generative artificial intelligence (AI) on key dimensions of financial market dynamics, namely volatility, liquidity, and return predictability. It emphasizes the dual role of AI as both an enhancer of informational efficiency and a potential source of short-term market instability. Methodology- The paper proposes an original AI-based sentiment indicator constructed using a hybrid large language model (LLM) framework that combines FinBERT and a GPT-4–class generative model. This sentiment index is analyzed alongside daily returns of major equity indices—S&P 500, NASDAQ, and STOXX 600—over the period 2018–2024. The empirical analysis relies on Principal Component Analysis (PCA), GARCH (1,1)-X models, and a Vector Autoregression (VAR) framework incorporating Amihud’s illiquidity measure, impulse response functions (IRFs), and forecast error variance decomposition (FEVD). Findings- Results from the GARCH-X estimations indicate that AI-driven sentiment is a statistically significant determinant of conditional volatility in U.S. equity markets. VAR-based Granger causality tests reveal a bidirectional relationship between AI sentiment and market returns, with particularly strong predictive effects for the S&P 500 and NASDAQ. Positive sentiment shocks are associated with improved market liquidity, as reflected by declines in the Amihud illiquidity ratio, while European markets display slower and weaker responses relative to U.S. markets. Conclusion- Generative AI functions as a double-edged mechanism in financial markets: it accelerates information processing and enhances short-term predictability, yet it may also amplify transient volatility through synchronized sentiment effects. Although AI has not fundamentally altered long-term market structures, its growing influence calls for renewed regulatory attention to AI-generated information flows and their implications for market stability Keywords: Generative artificial intelligence, market volatility, AI sentiment, algorithmic herding, large language models.