Abstract: Studying the construction of ChatGPT's discourse on China is conducive to revealing the ideological tendencies hidden in its discourse and the power relationships behind them. Representative keywords related to "socialism with Chinese characteristics" are adopted to interact with ChatGPT. Based on this, it uses Fairclough's three-dimensional model of critical discourse analysis as the theoretical framework and combines corpus technology to conduct discourse analysis on the text generated by ChatGPT. At the textual dimension, ChatGPT reproduces ideological bias against China under the dominance of Western values by constructing narratives such as authoritarian discourse, human rights discourse, and threatening conflict discourse. In the dimension of discourse practice, ChatGPT constructs irresponsibility through a combination of indirect discourse and unmarked discourse, while using Hegelian dialectical logic to present and summarize both positive and negative aspects. While strengthening the persuasiveness of discourse, it conceals the ideological stance behind it. From the perspective of social practice, the construction of ChatGPT's discourse on China is influenced by Western dominated datasets, intelligent discourse production logic, and the combined influence of American and Western political and capital ideologies. Research has shown that ChatGPT has not been able to break away from the ideological discourse framework of the United States and the West, and has become an agent of American and Western ideology under the cover of intelligent technology.