Large Language Models (LLMs) and low-code development platforms (LCDPs) have shown potential to fundamentally change the way how software products and applications are developed and to both combat the shortage of skilled software developers by increasing developer efficiency and lower the entry barrier for citizen developers to develop and maintain software on their own. This paper investigates whether these technologies can be combined to be even more powerful. To do this, a solution concept and prototype implementation were developed. Using the prototype, one can describe desired changes in an Oracle Application Express (APEX) low-code application to the GPT-4 Turbo LLM in a chat. The LLM then performs the changes for the user by calling specific edit functions that the system offers and generates replies to the user. This solution enables citizen developers to edit a fully functional Web application through natural language without help from a professional software developer. We also present a qualitative user study that we performed with ten Oracle APEX customers. It showed that participants have a rather positive opinion of both the fundamental concept and the prototype, liking aspects like its time-savings and ease of use. The study also uncovered some problems like a lack of a common vocabulary or technical understanding between the LLM and some users. However, participants already suggested ways to remedy such problems like integrating an element inspector into the prototype. Overall, the study highlights the feasibility and potential of the system and outlines multiple directions for its further development.