Have model transformation languages failed?

Abstract Model transformation plays an essential role in model-driven engineering. Models need to be transformed into other languages (e.g. for analysis) or rewritten (e.g. for optimisation or simulation). Specialised languages like ATL, ETL or QVT were designed for these tasks and became popular decades ago. However, looking at trends in academic venues and open-source repositories, we observe that their popularity has decayed in the last few years. This paper provides a retrospective on model transformation languages, analyses their current state in scientific venues and development platforms, and discusses possible reasons for their recent decline in popularity. Moreover, in light of current trends in AI-assisted development, we outline new research directions where model transformation languages can play a pivotal role, paving the way for their renewed adoption.