Using natural language to generate visual representations of data (NL2VIS) is emerging as a promising research direction, driven by the rapid development of Generative AI (GenAI). Given the growing number of studies in this area, this paper conducts a systematic review using the PRISMA method. Based on the analysis of 46 papers published between 2018 and 2024, the findings indicate that: 1) The number of publi-cations on this topic is increasing, with the IEEE Transactions on Visualization and Computer Graphics being the primary outlet; the Generative Adversarial Network (GAN) model serves as a foundational technology; 2) Research primarily focuses on integrating large language models (LLMs), such as ChatGPT, and enhancing the accuracy and interpretability of AI systems; 3) Current challenges include the lack of high-quality training data, limited transparency and interpretability, and ethical concerns related to the application of GenAI. Fu-ture research should aim to improve generative AI models, design user-centred interfaces, and address issues such as bias, trust, and privacy in AI-generated data visualizations