We construct a proxy for media opinion divergence using Latent Dirichlet Allocation. With this proxy, we investigate the impact of media opinion divergence on Chinese stocks. Findings indicate that higher media opinion divergence results in higher stock returns, but leads to lower future stock returns, consistent with Miller (1987). Additionally, we similarly explore the impact of divergence between traditional media and new media on stock returns, obtaining similar results. Findings are robust to controlling for firm characteristics and the number of news articles. Our study provides valuable insights into how the media can influence investor opinions and in turn the stock market in the digital media era.