<p>Understanding speech in noisy environments remains a persistent challenge for hearing aid users. Despite advances in amplification and digital signal processing, this limitation continues to impact communication and user experience. This review focuses on recent progress in artificial intelligence (AI) driven speech enhancement techniques designed for hearing aids, covering developments from 2017 to 2025. A systematic literature search across major databases was conducted using targeted keywords linking speech enhancement, hearing aids, and machine learning (ML). Following PRISMA guidelines, relevant studies were screened and categorized into three main groups: traditional signal processing methods, deep learning-based approaches, and hybrid models integrating artificial intelligence with adaptive filtering or contextual noise awareness. Findings show that deep learning techniques consistently outperform classical methods in complex acoustic scenarios. However, hybrid models offer favorable trade-offs in terms of latency and computational efficiency, which are key for real-time deployment in wearable devices. The review underscores a growing shift toward intelligent, personalized, and energy-efficient speech enhancement. It also highlights future directions in context-aware and on-device artificial intelligence solutions for assistive hearing technologies.</p>