This paper examines the information content of news media for the cross-section of expected equity option returns. We derive text-based signals from news articles on publicly traded companies that strongly forecast their delta-hedged equity option returns. The option return predictability of our textual signals is robust to variations in text representations and machine learning approaches. We propose a text-based method to evaluate various underlying mechanisms. We find that media coverage of companies contains valuable information about future change in stock return volatilities. This appears to be the key source of option return predictability by news articles.