ScholarWorks@숙명여자대학교

초록

This study empirically investigates the drivers of global market performance and localization strategies for leading generative AI applications (ChatGPT, Gemini, Claude, Grok, and DeepSeek) using iOS big data from 20 countries between 2023 and 2025. Methodologically, the research integrates quantitative performance with qualitative user responses through a two-stage panel analysis based on a fixed-effects model—controlling for country-specific invariant characteristics—and advanced BERTopic modeling integrated with Llama 3.1 and Mistral-Nemo models. The empirical results indicate that the volume of user evaluations (review counts) exerts a strong positive (+) influence on downloads and revenue. However, a "Rating Paradox" was observed, where market performance and average ratings conflict due to the coexistence of rapid user influx and initial dissatisfaction during stages of disruptive technological advancement. Notably, the panel analysis categorizes the global market into four distinct types—Signal-Sensitive, Qualitative Value-Selective, Disruptive/Mature Paradoxical, and Market Inertial—demonstrating the heterogeneity of regional adoption patterns. Furthermore, the topic modeling classified user interests into six domains: Intellectual Value, Creative Tasks, System Stability, Operational Efficiency, Platform Ecosystem, and Meta-feedback, identifying emotional bonding and infrastructure reliability as pivotal user experience factors. By proposing emotional localization for high-context cultures and Small Language Model strategies for emerging markets, this study provides strategic guidelines for generative AI firms to achieve successful "Glocalization."

키워드