As Large Language Models (LLM) emerge, opportunities for personalised learning are opening in education. LLMs are a valuable educational tool but raise ethical concerns regarding data privacy, consent, and potential bias reinforcement in higher education. Overusing AI-generated content can compromise critical thinking and problem-solving skills, resulting in less authentic learning. This creates a big question about using this disruptive model in education. To understand this, we attempt to review the use of LLMs in higher education through a systematic literature review utilising PRISMA approach . The findings reveal crucial insights into several challenges, but researchers are inclined to embrace LLMs for their benefits. The analysis reveals that higher education institutions must establish robust academic integrity policies and ensure AI-based assessments do not solely determine learning outcomes. A balanced approach that values social interaction, reflection, and collaboration is key to success in the AI world.