This article presents a case study examining the use of large language models (LLMs) in the process of creating an interactive map of Krakow’s digital cultural heritage. The aim of the study was to analyse the potential of LLMs as tools supporting fast prototyping of web maps in a popularisation and narrative context. The research methodology consisted of comparing two variants of the generated map using differently configured language models, while maintaining the same point dataset and the Leaflet.js programming library. In both variants, the LLM served as a client- side application code generator, and a tool supporting the integration of descriptive data and user interface design. The results suggest that a more elaborate user interface was generated using the standard ChatGPT interface, whilst the specialised GPT model produced a solution with lower logical complexity, understood as the range of decision-making rules and data processing implemented on the client side of the application. This confirms that the degree of language model specialisation does not translate directly into the functional scope of the generated map application. The conclusions highlight the role of LLMs as tools that support the design process and shorten the prototyping cycle in web cartography, while maintaining expert control over the data and their interpretation.