The paper presents a framework for detecting social bots on Twitter, leveraging over a thousand features extracted from user metadata, tweet content, network patterns, and activity time series. The framework is evaluated using a publicly available dataset of Twitter bots and a manually annotated collection of active Twitter users. The models achieve high accuracy and can detect bots of different sophistication levels. The study estimates that between 9% and 15% of active Twitter accounts are bots. The analysis of user interactions reveals that humans tend to interact with more human-like accounts, while simple bots interact with more sophisticated bots. Clustering analysis identifies several subclasses of accounts, including spammers, self-promoters, and accounts that post content from connected applications. The paper also discusses the evolution of bot sophistication and the challenges in detecting increasingly sophisticated bots.The paper presents a framework for detecting social bots on Twitter, leveraging over a thousand features extracted from user metadata, tweet content, network patterns, and activity time series. The framework is evaluated using a publicly available dataset of Twitter bots and a manually annotated collection of active Twitter users. The models achieve high accuracy and can detect bots of different sophistication levels. The study estimates that between 9% and 15% of active Twitter accounts are bots. The analysis of user interactions reveals that humans tend to interact with more human-like accounts, while simple bots interact with more sophisticated bots. Clustering analysis identifies several subclasses of accounts, including spammers, self-promoters, and accounts that post content from connected applications. The paper also discusses the evolution of bot sophistication and the challenges in detecting increasingly sophisticated bots.