As ChatGPT and the underlying Large Language Models are rapidly gaining popularity, artificial intelligence-generated content (AIGC) receives wide attention both from the academic and the industry. However, existing research has not fully explored the detection methods. With 47,151 Weibo marketing real cases from 1l,610 creators on Weibo, this paper creates AIGC cases using ChatGLM and conducts linguistic analysis, including vocabulary features, part-of-speech analysis, sentiment analysis, language concreteness, and topic modeling. Besides, this paper adopts current methods (including AdaBoost, Random Forest, GBDT, XG-Boost, SVM, StackingClassifer, and RoBERTa) to detect AIGC and finds that RoBERTa outperforms other models with the comparison dataset. This paper analyzes and compares the features of human- and AI-generated marketing content and provides effective detection methods.