Analysis of Deepfake using Large language Model

Deepfakes have evolved from an experimental curiosity into a pervasive threat to trust, financial integrity, and personal dignity. While classical detectors continue to offer value, they frequently fail to identify emerging forms of deepfakes and provide minimal justification for their judgments. LLMs introduce semantic comprehension that can be leveraged for explanatory purposes. Based on this review, the capabilities offered by LLMs are genuine but constrained. As zero-shot judges, they facilitate triage but cannot independently render a definitive judgment. When subjected to fine-tuning with language supervision, such models demonstrate acceptable levels of accuracy, localization, and explainability. Performance metrics for audio detection remain inconsistent even in state-of-the-art configurations, and for text detection, the core problem is inherently probabilistic. Both technical findings and the Arup incident underscore the need for a layered approach: provenance verification and labeling, specialized detectors, LLM-driven reasoning, and human verification operating collaboratively, without exclusive reliance on any singular indicator.  For example, India's 2026 amendments establish transparency as a regulatory principle, and the practical implementation of these regulations will ultimately determine their efficacy.  In summary, LLMs are powerful adjuncts in combating deepfakes, but they are not infallible truth-tellers.

[1]  A. Rössler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies and M. Nießner, “FaceForensics++: Learning to detect manipulated facial images,” in Proc. IEEE/CVF Int. Conf. Computer Vision (ICCV), 2019. [2]   Y. Li, X. Yang, P. Sun, H. Qi and S. Lyu, “Celeb-DF: A large-scale challenging dataset for DeepFake forensics,” in Proc. IEEE/CVF Conf. Computer Vision and Pattern Recognition (CVPR), 2020, pp. 3207-3216. [3]  Y. Li et al., “Celeb-DF++: A large-scale challenging video DeepFake benchmark for generalizable                              forensics,” arXiv:2507.18015, 2025. [4] T. T. Nguyen et al., “Deep learning for deepfakes creation and detection: A survey,” arXiv:1909.11573. [5]  T. Wang, X. Liao, K. P. Chow, X. Lin and Y. Wang, “Deepfake detection: A comprehensive survey from the reliability perspective,” arXiv:2211.10881. [6]  G. Pei et al., “Deepfake generation and detection: A benchmark and survey,” arXiv:2403.17881, 2024. [7] “Exploring autonomous methods for deepfake detection: A detailed survey on techniques and evaluation,” review article, ScienceDirect,      2025  (PII S240584402500653X). Deepfakes and Their Analysis Using Large Language Models Page 11 [8]    S. Jia et al., “Can ChatGPT detect DeepFakes? A study of using multimodal large language models for media forensics,” in Proc. CVPR Workshops (WMF), 2024. [ [9]   Z. Yang et al., “The dawn of LMMs: Preliminary explorations with GPT-4V(vision),” arXiv:2309.17421, 2023. [10]   “A review of large language models for multimodal deepfake detection,” preprint, ResearchGate, 2025.