Abstract: The rapid emergence of generative artificial intelligence (AI) systems and large language models has transformed the ways users interact with digital technologies, creating new challenges and opportunities for user experience (UX) research. Despite the growing adoption of AI-powered conversational systems, the intellectual structure, thematic evolution, and emerging research directions of UX studies in this domain remain insufficiently understood. Therefore, this study aims to analyse the evolution of UX research in generative and conversational AI systems, identify its major thematic areas, and uncover emerging research trends and future directions. A set of 3,714 publications published between 2006 and 2026 was analysed using bibliometric techniques and topic modelling. The dataset was constructed through a systematic search of the Scopus database, resulting in 3,714 eligible records for analysis. The records were pre-processed using standard text-mining procedures, including normalisation, tokenisation, lemmatisation, and keyword integration, to ensure data quality and analytical consistency. Temporal evolution was examined through descriptive statistics, exponential moving averages, and change-point analysis. The findings reveal a major transformation in the field, characterised by the rapid expansion of research after 2023 and the diversification of UX-related themes across multiple application domains. To identify latent thematic structures, Non-negative Matrix Factorisation (NMF), a topic modelling technique that decomposes document–term matrices into interpretable topic representations, was employed. Non-negative Matrix Factorisation (NMF) topic modelling and keyword co-occurrence network analysis identified several dominant research themes, including intelligent software and product development, education and e-learning systems, retrieval-augmented generation (RAG)-based information systems, e-commerce and customer experience, immersive and extended reality environments, usability evaluation, and emotional user interaction. The co-occurrence network further revealed well-defined intellectual clusters centred on healthcare, educational technology, social computing, and intelligent interactive systems. Emerging trends were identified through a multi-signal approach combining discriminative terms, burst analysis, and growing topic trajectories. Key emerging concepts included retrieval-augmented generation (RAG), multimodality, agent-based systems, alignment, trust, privacy, hallucination, and explainability. Overall, the findings demonstrate a shift from conventional UX concerns toward increasingly adaptive, multimodal, trustworthy, and human-centred AI ecosystems. The study provides a comprehensive map of the field’s evolution and highlights future research priorities for designing transparent, explainable, and collaborative AI-powered user experiences.