Choosing What Works: Performance-Based Trust in LLMs Among Chinese International Students

This paper explores how Chinese international students in the United States navigate trust, identity, and political boundaries when using large language models (LLMs). Drawing on interviews with fourteen students, I examine how they evaluate AI systems such as ChatGPT and DeepSeek across different linguistic and informational environments. While discussions of AI often emphasize nationalism and geopolitical competition, participants rarely based their trust on a model's national origin. Instead, they prioritized performance, accuracy, and task-specific usefulness. The study argues that trust in AI is shaped less by national affiliation than by everyday practices of comparison, verification, and pragmatic evaluation.