Failure Mode and Effects Analysis(FMEA) is a central tool in risk management, particularly in highly regulated industries. The creation of an FMEA is time-consuming but enables early detection of potential errors and identification of critical risks, which ensures the safety of the production process and compliance with quality standards. Despite its importance, automation of FMEA processes remains limited, primarily due to the complexity of analyzing or generating language-based content. Large Language Models (LLMs) offer a solution for generating content in natural language, which can support and potentially improve FMEA creation. An important aspect of using LLMs is the design of the prompt, which influences the quality and accuracy of the results generated. This paper investigates how different prompts affect the quality of FMEA content generated by LLMs. By analyzing various prompt design strategies in different stages of FMEA creation, this study quantitatively and qualitatively evaluates the impact on the quality of generated FMEAs. These insights highlight the potential of LLMs for the creation of FMEAs in industry and can be used to improve the efficiency and consistency in FMEA creation.