Abstract
This paper examines how algorithmic trading (AT) affects limit order book (LOB) structure using agent-based simulations. I use two complementary settings: a controlled simulation that varies the share of AT directly, and a trade-driven simulation calibrated to empirical order flow for Jet Blue and Google. Across both settings, AT generally improves market quality when the market is stable by narrowing spreads, increasing liquidity, and reducing short-horizon return dispersion and volatility. However, it may amplify certain distorted dimensions, if present. The magnitude and clarity of effects also depend on market conditions.