The rise of social media has brought a wave of automated programs known as social media bots. These bots pose significant challenges by disseminating misinformation, shaping fake trends, and manipulating public opinion. This also leads to a loss of trust in the platform and creates an environment of suspicion. The advent of the Large Language Model has further exacerbated this issue, as bots now generate content that is increasingly indistinguishable from human posts, making bot detection more difficult. Despite extensive research in social media bot detection, very few studies specifically focus on bots leveraging Generative AI technology. To bridge this research gap, in this paper, we introduce a novel enhanced version of the BotRGCN model, called Intent-Spectrum BotTracker, by incorporating three innovative features: user intention, posting message topic, and posting message entropy. Extensive experiments have been conducted, and the results demonstrate that the integration of intention, topics, and entropy metrics significantly enhances the performance of various baseline models, with our enhanced BotRGCN model exhibiting the most superior social media detection capabilities.