This paper explores the potential of Generative Artificial Intelligence (GAI) tools in automatically converting complex texts in Spanish language into more accessible forms, in alignment with the Spanish Plain Language regulations. For this study specifically, the DeepSeek tool was analyzed. Two strategies have been implemented: the first strategy employs the default model of DeepSeek guided by optimized prompts; the second strategy implements a full pipeline involving HTML and metadata preprocessing, fine-tuning of the ByT5-small model of DeepSeek, and subsequent plain language generation. To verify compliance with plain language regulations for the new texts generated using both strategies, two classic Spanish text readability measures were used, along with a morphosyntactic analysis of the texts. The study used a dedicated corpus based on Spanish National Administration web pages to fine-tune the DeepSeek model. Experimental results show that, although the prompting strategy yields excellent values on different readability indices, the fine-tuned model provides more stable and consistent improvements.