The detection of human trafficking remains a complex and urgent challenge, further compounded by the adaptive strategies employed by traffickers on online platforms. We propose a novel architectural model that integrates generative artificial intelligence (AI) with social science insights for the effective identification of human trafficking activities. The proposed software solution leverages advanced AI techniques, including computer vision, natural language processing (NLP), Large Language Models (LLMs) and deep neural networks (DNNs), to analyze web-based text-image data. Key components include biometric analysis via DeepFace and few-shot text classification, supported by synthetic data generation aligned with the Department of Homeland Security (DHS) Strategic Plan. To enhance performance, various vision-language models, detectors, and analytical methods were compared and integrated into a fusion-based tool capable of evaluating the likelihood that an online escort advertisement aligns with patterns indicative of human trafficking. Preliminary results demonstrate a notable accuracy rate exceeding 90% in identifying possible trafficking-related advertisements, underscoring the model’s potential to significantly improve detection capabilities. Ongoing efforts focus on refining model precision and developing realistic synthetic image data to mitigate data scarcity challenges while maintaining ethical considerations. This interdisciplinary approach advances existing tools for homeland defense and law enforcement while prioritizing a victim-centered strategy. By leveraging AI-driven technologies, the proposed solution offers transformative potential for combating human trafficking and safeguarding vulnerable populations.