This thesis investigates the potential of large language models (LLMs) in the automatic simplification of texts into Easy Italian. Two advanced models, ChatGPT (OpenAI) and Claude Sonnet (Anthropic), were used to simplify two Swiss institutional texts through different prompting strategies, including zero-shot, prompt chaining, and fine-tuning. Sixteen outputs were produced and evaluated through a combined approach of qualitative assessment and automatic analysis.The results highlight the central role of prompting strategies and the importance of combining automatic analysis with human evaluation to ensure that the texts are genuinely inclusive.By presenting a reproducible evaluation framework and offering practical recommendations, this research contributes to the field of AI-assisted text simplification and provides insights for institutions and public services seeking to promote more inclusive communication practices.