Current Research on Geography, Earth Science and Environment Vol. 9

The accelerated adoption of generative artificial intelligence (AI) tools in higher education has outpaced our understanding of the conditions under which they actually support technical learning. This chapter addresses that gap through an exploratory qualitative case study conducted with environmental engineering students at Universidad Santo Tomás (Colombia) during the second academic semester of 2025. The study examines how prior disciplinary knowledge in geographic information systems (GIS) shapes the way students mobilise freely available AI tools, ChatGPT and DeepSeek, when facing an authentic geospatial task: the multitemporal analysis of Lake Tota using the Normalized Difference Water Index (NDWI) in ArcGIS Pro. Drawing on a corpus of queries submitted to the AI tools by two contrasting groups, the study documents a clear pattern. Students without a conceptual baseline produced generic queries, reached the central concept of the task only near the end of their process, could not interpret the model's responses, and abandoned the task amid frustration; none reached a verifiable result. Students with prior GIS training anchored their very first query in the technical concept, recovered from errors within minutes, and most reached verifiable outcomes, although a recurring error in the Landsat band formula revealed the risks of uncritical trust in AI output. Interpreted through the lenses of meaningful learning, AI literacy, and cognitive offloading, the findings suggest that AI amplifies existing disciplinary competence rather than substituting for it, and that the instructor remains decisive in the conceptual scaffolding stage. The chapter contributes a transferable reflection for integrating AI into technical curricula in resource-constrained contexts.