This paper examines the relationship between investor sentiment and the volatility of China's stock markets. We use the data from Baidu, China's leading search engine, for information on investor sentiment. In two different Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models, the baseline model and the Baidu-Index extended model, we forecast the return volatility of China's stock markets. We find that the Baidu-Index extended model performs better than the baseline model. This paper shows that the search volume of relevant key words from Baidu Index improves volatility forecasting of China's stock markets.