Social media platforms have become powerful spaces for spreading extremist ideologies, with several violent attacks in the United States linked to online radicalization processes. Artificial Intelligence (AI) technologies, including machine learning and natural language processing, offer potential solutions for detecting extremist content and identifying radicalization pathways. This scoping review examines how AI has been utilized to identify online radicalization pathways in the U.S. context, synthesizing research across computer science, criminology, and security studies. A systematic search of electronic databases and grey literature sources was conducted following established scoping review methodology. The review reveals that AI applications in radicalization detection primarily focus on content classification, user behavior analysis, and network mapping, with techniques that range from traditional machine learning to advanced deep learning models. However, significant challenges remain, including definitional inconsistencies, dataset limitations, ethical concerns about surveillance and privacy, and the need to balance public safety with constitutional protections. This review identifies critical gaps in current research and highlights the need for interdisciplinary collaboration to develop effective, ethical AI-based approaches to counter online radicalization while preserving civil liberties.