The advancement of generative models has enabled the creation of synthetic faces with a high degree of realism, posing challenges in verifying their authenticity. This study proposes an approach based on the analysis of eye features to detect artificially generated face images. To achieve this, three key aspects are examined: pupil shape, iris color, and the similarity of corneal specular reflections. Additionally, a solution is proposed to improve the segmentation of the iris by connecting edges in masks with holes using the YoloV8 network, addressing a common challenge in semantic segmentation. The results show that the proposed strategy performs well, achieving an accuracy exceeding 80% in identifying artificially generated faces.