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 have
emerged as transformative technologies in education, yet comprehensive comparative
analysis of their pedagogical applications and effectiveness remains limited. Despite
widespread adoption and growing interest, critical research gaps persist regarding their
implementation strategies, educational outcomes, and ethical considerations across di
verse learning contexts. To address these gaps, we conducted a systematic review fol
lowing PRISMA guidelines across seven major academic databases, synthesizing find
ings from empirical studies on LLM integration in formal and informal educational set
tings. The analysis identified distinct pedagogical affordances among the four models:
ChatGPT demonstrated superior conversational learning capabilities and general knowl
edge support, DeepSeek exhibited exceptional performance in programming education
and technical domains, Gemini led in multimodal educational applications particularly
for STEM subjects, and Qwen showed enhanced multilingual competency and cultural
sensitivity for diverse international contexts. Our findings revealed significant dispari
ties in adoption patterns, with higher education demonstrating greater integration success
compared to K-12 environments, which face substantial institutional and ethical barri
ers. Key challenges identified include academic integrity concerns, over-reliance risks,
limited longitudinal impact evidence, systematic underutilization of multimodal capabil
ities, and integration difficulties with existing Learning Management Systems. Addition
ally, we identified critical gaps in personalized learning mechanisms, cultural adaptation
frameworks, and comprehensive ethical guidelines. The study provides evidence-based
recommendations for optimal LLM selection and implementation, proposes frameworks
for addressing identified challenges, and establishes a foundation for future empirical re
search. As educational AI continues evolving rapidly, these findings serve as an essential
reference for educators, researchers, and policymakers to leverage LLM strengths while
addressing limitations and ethical considerations in authentic educational contexts.