Deep fake technology poses a significant threat to the authenticity of digital content, necessitating the development of robust detection mechanisms. In this research, we propose a novel approach that combines Deep Q-Learning (DQL) and Artificial Gorilla Troop Optimization (AGTO) to achieve unprecedented levels of accuracy in deep fake detection, with a remarkable 99% accuracy rate. Our methodology utilizes DQL to learn optimal detection strategies by framing the decision-making process as a reinforcement learning problem, enabling the agent to discern subtle patterns indicative of deep fake manipulation. Additionally, we introduce AGTO, inspired by the collaborative behaviour of gorilla troops, to enhance the optimization process of feature selection and model tuning. Through extensive experiments on diverse deep fake datasets encompassing various manipulation techniques and levels of sophistication, we demonstrate the superior performance of our approach, consistently achieving a detection accuracy of 99% across different scenarios