This paper investigates whether recent developments in machine translation (MT) have affected two well-documented stylistic characteristics of MT: reduced lexical diversity and increased source syntax mirroring. Using a corpus of English into French translations, we compared MT outputs produced by a widely used system over time (2024 vs. 2026) and under different models (NMT vs. LLM-based). Lexical diversity is measured using word translation entropy (HTra), and source syntax mirroring using four syntactic similarity metrics. Results show that lexical and syntactic characteristics of NMT output remain stable over time. The LLM-based MT exhibits a slight increase in lexical diversity, but still presents strong signs of lexical overgeneralisation. Syntactic similarity to the source remains largely unchanged. Overall, the findings suggest that recent MT developments have a very limited impact on these two stylistic characteristics.