The expansion of artificial intelligence (AI) and machine learning (ML) in urban planning and smart city studies extends beyond the development of analytical tools and is accompanied by changes in knowledge production, urban problem definition, and decision support. Despite the rapid growth of research in this field, the relationship between thematic transformations in the literature and changes in the logic of urban knowledge production has not yet been systematically examined. This study investigated thematic transformations and epistemological shifts in the AI and ML literature on urban planning and smart city studies during 2020–2025. The novelty of the study lies in integrating bibliometric analysis with epistemological interpretation to explain the reconfiguration of the field’s knowledge structure. Data were retrieved from the Web of Science and Scopus databases, and after systematic screening based on the PRISMA 2020 framework, 3,578 articles were included. Using a scientometric–interpretive approach, the study applied thesaurus-based supervised thematic classification, temporal trend analysis, the Smoothed Relative Growth Index (RGI), epistemological paradigm analysis, scientific collaboration analysis, and geographical analysis of knowledge production. The results showed that Urban Environment and Sustainability was the largest thematic domain (31.9%). At the epistemological level, the data-driven and predictive paradigm dominated knowledge production (77.7%), while generative AI approaches are gradually expanding. Knowledge production was also found to be interdisciplinary but geographically concentrated in a limited number of countries. Overall, the findings indicate an intra-disciplinary reconfiguration in the urban AI literature, with AI evolving from a set of analytical tools into an infrastructure for knowledge production, scenario generation, and urban decision support. They also highlight the need to address data governance, transparency, equity, and inequalities in knowledge production to support the future development of urban AI.