The proliferation of synthetic media generated through deep learning architectures — commonly termed deepfakes — constitutes a growing threat to information integrity, personal security, and democratic processes. This research investigates the effectiveness of artificial intelligence-driven frameworks in detecting deepfake content across visual and multimodal domains. Employing a descriptive-analytical design, primary data were systematically collected from 120 respondents comprising cybersecurity professionals, AI engineers, academic researchers, and media forensics specialists through a structured Likert-scale survey instrument. Three hypotheses were formulated and tested using one-sample t-tests at a 95% confidence level. Results demonstrate that AI-based ensemble models — particularly those combining Vision Transformer and CNN architectures — achieved an F1-score of 94.9%, significantly outperforming single-architecture baselines. Findings further confirm strong professional consensus on the effectiveness of detection technologies and a statistically significant relationship between public awareness and adoption intent. The paper concludes with policy recommendations, investment priorities, and directions for future research in adversarial deepfake mitigation.
Vaishnavi Ghanshyam Vyas (Fri,) studied this question.