Software requirements specifications (SRS) are essential to the success of software development. It is widely recognized that the quality of SRSs affects both the quality of their product and project results. IEEE 830 and ISO/IEC/IEEE 29148 are the international standards for the requirements process and SRSs. These standards provide recommended section structures and content of SRSs. However, a study has shown that many organizations have not adopted these standards due to a lack of knowledge or cost constraints. Indeed, there is a wide variety of SRS structures. This paper explores an approach for automatically converting SRSs with any structures into the standard structure. It is considered beneficial to standardize SRS structures. A key benefit is that it facilitates research on SRSs. It makes the assumption that SRSs in research follow the standard structure more realistic. In fact, some studies have assumed that sections of a SRS correspond to those of the standard structure. To achieve SRS structural standardization, we use a large language model (LLM) that demonstrates high natural language processing (NLP) capability. We designed simple prompts and assessed the feasibility of an LLM. As a result, standardized SRSs contained about 80% of the required items in the standard structure, but they could not fully retain all information from the original SRSs. Observations of the results suggest that prompts should be designed to consider the size equivalence before and after standardization and relations between sections.