In this paper, we introduce an innovative improvement to the traditional Extended Kalman Filter-based Simultaneous Localization and Mapping framework, called Entropy-Gated Feature-Aware EKF-SLAM. Our approach presents a flexible information-theoretic mechanism that adjusts the impact of each landmark observation in real-time, depending on the historical entropy of its measurement distribution. Traditional EKF-SLAM assumes that all observations are equally trustworthy based on noise models. In contrast, our approach assesses long-term measurement consistency using rolling entropy profiles, which allows for per-landmark trust gating during the innovation update. Each landmark's innovation is specifically weighted by a coefficient derived from normalized entropy, which enables the filter to discount features that are ambiguous, noisy, or aliased, while placing emphasis on landmarks that are historically stable and informative. This entropy-aware gate is integrated directly into the EKF correction equations, maintaining the prediction model intact and ensuring computational efficiency is preserved. A complete mathematical derivation supports theoretical development, and we present a detailed simulation in MATLAB featuring a mobile robot navigating a 2D environment with five landmarks. Results from experiments show enhanced accuracy in trajectory and greater robustness when faced with perceptual ambiguity and dynamic sensor noise. This change represents an important advancement in adaptive uncertainty modeling in filtering-based SLAM frameworks.