Over the last decade, controlled languages (CL) have received increased attention in machine translation (MT) research. The vast majority of studies have dealt with the impact of CLs on the quality of the final MT output, but very little work has focused on the impact of MT on the accessibility of target texts for people with special needs. This article represents a first attempt to bridge this gap. We present a comparative linguistic study that seeks to explore whether MT systems are a viable option for translating texts that are easy to read and understand (EtR). We tested DeepL, Google Translate, and Yandex with EtR texts from three different domains in four language pairs. Findings show that DeepL is the highest-performing system, and that Spanish and administrative texts in particular seem to present more challenges. The evaluation of the MT output in terms of linguistic accessibility indicates that the highest number of issues are found at a lexical and stylistic level. Although MT systems do not generate EtR texts of acceptable quality yet, our study highlights the potential of this tool, as well as the challenges of creating multilingual content that is accessible for all.