<p>Large language models e.g. ChatGPT facilitate users and are emerging as important customer interaction interfaces. However, little research has explored how ChatGPT-generated content source disclosure (i.e., disclosing the generative origin such as webpages, etc.) shape user responses. Based on signal theory, this article explores the effect of ChatGPT-generated content source disclosure (vs. non-disclosure) on user responses. The study conducted three online scenario experiments and collected 500 respondents with usage experience from a survey platform. Through three studies, we reveal that ChatGPT-generated content source disclosure (vs. non-disclosure) significantly increases user satisfaction, usage intention, and word-of-mouth, with this effect mediated by perceived competence. Moreover, ChatGPT-generated content source disclosure is more effective in increasing users’ perceived competence in subjective (vs. objective) task situations. This article provides new insights into understanding the characteristics of ChatGPT-generated content and optimizing the user experience.</p>