A Survey on Large Language Models for Job Recommendations
Sarang Shelke, Aditya Shelke, Vedant Lodha & Yashraj Gaikwad
Department of Artificial Intelligence and Data Science
PVG’s COET, Pune, India
Abstract—The rise of Large Language Models (LLMs) has transformed job recommendation systems by moving beyond traditional keyword-based matching and collaborative filtering to a more context-aware and intelligent approach. Leveraging deep contextual understanding and external knowledge, LLMs can analyze job descriptions and candidate profiles with greater accuracy, enabling more personalized and meaningful job rec-ommendations. Techniques such as fine-tuning and prompt engi-neering enhance their ability to establish complex relationships between candidates and job roles, improving overall matching quality. However, despite their potential, LLM-powered job recommendation systems face challenges such as bias, explain-ability, and scalability. Many existing models function as ”black boxes,” offering little transparency in how recommendations are generated, making it difficult for job seekers to interpret AI-driven career guidance. Furthermore, conventional methods are often restricted to retrieving and ranking job postings rather than acting as dynamic career assistants that proactively provide tailored suggestions. This survey categorizes existing LLM-based job recommendation approaches into discriminative and genera-tive paradigms, analyzing key methodologies, comparing different architectures, and identifying their strengths and limitations. We also discuss recent advancements, address ongoing challenges, and explore future research directions to enhance fairness, transparency, and adaptability in AI-driven hiring solutions.
Index Terms—Job Recommendation, Large Language Models, Context-Aware Matching, NLP, AI in Recruitment.
