Дослідження можливостей ші чат-ботів для виявлення фейкового відеоконтенту

Abstract

This study investigates the effectiveness of using general-purpose Large Language Model (LLM) chatbots for detecting fake video content. The relevance of this topic is driven by the rapid spread of misinformation on social media, which necessitates the development of accessible tools for rapid fact-checking. Throughout the research, a dataset was compiled consisting of 10 verified fake videos and 10 authentic materials. The study analyzes the ability of various AI models (including Gemini and its counterparts) to identify signs of deepfakes, manipulative audio, and visual inconsistencies. The research methodology involved segmenting long video clips to optimize processing within the models' input data limits. The results demonstrate the accuracy and reliability of chatbots as tools for initial forgery detection.