Phishing continues to be a major issue not only affecting the internet users but also being a big problem for the cybersecurity world; university networks are the most probable attack targets due to their open infrastructures, diverse user population, and limited security resources. The breakthrough in artificial intelligence (AI), notably large language models, has not only made phishing attacks more sophisticated and realistic but also envisioning new defense techniques based on machine learning and natural language processing. This current research report is a systematic literature review (SLR) of 53 academic studies that examine the dual aspect of AI in promoting and hindering phishing attacks in higher education institutions (HEIs). The review reveals three prominent points: emails sent by AI are increasingly real and adaptable; AI-based detection systems are very effective in laboratory-like conditions but struggle against new and adversarial attacks; and human factors like lack of user awareness and slow incident reporting are still the main vulnerabilities. The research then proposes a multi-layered defense framework that includes infrastructure strengthening, AI detection, human-centered awareness training, incident response mechanisms, and governance policies. This framework provides a practical roadmap for HEIs to boost their cybersecurity resilience and play a part in the sustainable growth of the university.