This study aims to explore how users’ positive and negative perceptions of text-based generative artificial intelligence are manifested and to examine the latent factor structure underlying these perceptions. While prior technology acceptance research has primarily focused on functional benefits, text-based generative artificial intelligence exerts influence across information seeking, problem solving, and broader cognitive processes, thereby necessitating a more integrative and ambivalent analytical perspective. To address this, the present study adopted an exploratory sequential design. In the initial stage, in-depth interviews were conducted with 20 adult users of text-based generative artificial intelligence to identify key attributes of user perceptions. Subsequently, survey data were collected and analyzed using exploratory factor analysis. The results indicated that positive perceptions were categorized into four factors: task efficiency, performance reliability, information retrieval efficiency, and cognitive expansiveness. Negative perceptions, in contrast, were grouped into three factors: interaction inaccuracy, information distrust, and cognitive deterioration. These findings suggest that text-based generative artificial intelligence encompasses both functional benefits and potential risks, and that user perceptions are characterized by a complex and ambivalent structure rather than a single evaluative dimension. As an exploratory investigation, this study contributes by presenting a preliminary understanding of the structural dimensions of user perceptions toward text-based generative artificial intelligence, providing a starting point for subsequent theoretical discussions and future research. In addition, the findings may offer foundational insights for ongoing discussions related to technology design and user-oriented policy considerations.