Recent advances in conversational AI have created new opportunities for learning English. While these technologies show promising potential and have positively influenced English learning, they remain constrained by limited interaction channels and occasional output of irrelevant or undesired information. This systematic review analyzed 52 empirical studies of conversational AI published between 2019 and 2024, applying specific criteria to evaluate their effectiveness in supporting English learning. The latest models demonstrate state-of-the-art performance, offering potential solutions to existing limitations. This review provides future researchers with an overview of the potential applications of various models, offering insights to guide the development of next-generation conversational AI for English language learning.