The Impact of Large Language Models in Education: A Review of ChatGPT, DeepSeek, Gemini, and Qwen

Large language models (LLMs) including ChatGPT, DeepSeek, Gemini, and Qwen haveemerged as transformative technologies in education, yet comprehensive comparativeanalysis of their pedagogical applications and effectiveness remains limited. Despitewidespread adoption and growing interest, critical research gaps persist regarding theirimplementation strategies, educational outcomes, and ethical considerations across diverse learning contexts. To address these gaps, we conducted a systematic review following PRISMA guidelines across seven major academic databases, synthesizing findings from empirical studies on LLM integration in formal and informal educational settings. The analysis identified distinct pedagogical affordances among the four models:ChatGPT demonstrated superior conversational learning capabilities and general knowledge support, DeepSeek exhibited exceptional performance in programming educationand technical domains, Gemini led in multimodal educational applications particularlyfor STEM subjects, and Qwen showed enhanced multilingual competency and culturalsensitivity for diverse international contexts. Our findings revealed significant disparities in adoption patterns, with higher education demonstrating greater integration successcompared to K-12 environments, which face substantial institutional and ethical barriers. Key challenges identified include academic integrity concerns, over-reliance risks,limited longitudinal impact evidence, systematic underutilization of multimodal capabilities, and integration difficulties with existing Learning Management Systems. Additionally, we identified critical gaps in personalized learning mechanisms, cultural adaptationframeworks, and comprehensive ethical guidelines. The study provides evidence-basedrecommendations for optimal LLM selection and implementation, proposes frameworksfor addressing identified challenges, and establishes a foundation for future empirical research. As educational AI continues evolving rapidly, these findings serve as an essentialreference for educators, researchers, and policymakers to leverage LLM strengths whileaddressing limitations and ethical considerations in authentic educational contexts.