Access to information is essential for everyone to actively participate in society. Easy language, as the most comprehensible language variety, is a key instrument in making content accessible to a larger audience. Recently, since the boom of large language models, their potential as a tool to improve the accessibility of written content has been tested in various contexts. This study aims to explore the intersection between easy language and large language models by determining, primarily, whether these models produce the best results when translating interlingually or intralingually. To do so, it featured an expert evaluation (which assessed the models’ output following easy language guidelines and translation quality metrics), an end-user evaluation (including a questionnaire and a usability study), as well as some tests with automatic metrics. By answering our main research question, we concluded that interlingual translation performed the best, a statement that is supported by our different evaluation processes.