When AI feels more human: Language and user experiences with ChatGPT

Abstract

Purpose: This study examines how using a native versus a foreign language shapes users' perceptions of ChatGPT in terms of linguistic proficiency, cultural fluency, emotional connection, and perceived human likeness. Drawing on Media Equation Theory, the study explores how multilingual contexts influence human-AI interaction and users' interpretations of humanlike qualities in AI communication. Design/methodology/approach: Participants interacted with ChatGPT in both their native language (Turkish) and a foreign language (English). Following these interactions, semi-structured interviews lasting approximately 30-45 min were conducted, audio-recorded with participants' consent, and transcribed verbatim. The data were analyzed using Interpretative Phenomenological Analysis (IPA). Through an iterative coding process, three main themes and nine sub-themes were identified. Findings: Language plays a significant role in shaping perceptions of AI interaction. Participants with lower English proficiency attributed greater humanlike qualities to ChatGPT and reported a stronger emotional connection when interacting in Turkish. However, interactions in English were generally perceived as more efficient and intelligent. Participants also raised concerns regarding ChatGPT's cultural fluency, particularly when culturally specific expressions were not adequately reflected. Overall, the results suggest that multilingual users may experience a tension between emotional connection and perceived AI competence across languages. Research limitations/implications: The study involved 33 bilingual participants, and the findings may not be generalizable beyond this particular context. Nevertheless, the results highlight the importance of linguistic and cultural context in shaping human-AI interaction and suggest that multilingual environments may influence how users evaluate AI systems. Practical implications: Findings suggested a linguistic-emotional trade-off among non-native English-speaking users: participants commonly described Turkish-language interactions as fostering greater emotional connection and relational closeness, whereas English-language interactions were more often perceived as more competent and intelligent. The findings indicate that improving linguistic accuracy and cultural responsiveness in AI systems may enhance inclusivity, perceived empathy, and user experience in multilingual contexts. Originality/value: This study contributes to the growing literature on human-AI interaction by examining how multilingual users perceive AI systems across languages and by demonstrating how linguistic background shapes emotional and social perceptions of AI communication.