Wetlands are critical ecosystems for biodiversity, water security, carbon cycling, flood regulation, and climate adaptation; however, their multi-scale and dynamic character makes them difficult to communicate through conventional science education. This article integrates recent scholarship on environmental education, science communication, informal learning, extended reality (XR), and artificial intelligence (AI) to examine how AI-enhanced XR science exhibitions can support wetland education. Virtual, augmented, and mixed reality, digital twins, generative AI, intelligent virtual guides, and adaptive learning are synthesized through the lenses of multimodal learning, cognitive load, social cognitive theory, socio-scientific issues, and place-based ecopedagogy. Evidence suggests that XR is particularly valuable for spatial understanding, presence, engagement, and the visualization of otherwise inaccessible ecological processes, but immersion alone does not guarantee durable learning or pro-environmental attitude change. AI can add adaptive explanation, personalized feedback, multilingual dialogue, and responsive scenario design, while simultaneously amplifying concerns about scientific fidelity, algorithmic bias, privacy, and over-automation. As its main contribution, the article proposes the AI-XR Wetland Exhibition Framework (AIXR-WEF), which links local ecological data, immersive representation, AI adaptation, inquiry and decision-making, action pathways, and ethical governance. The framework offers design principles, evaluation indicators, and a research agenda for science centers, museums, visitor centers, and university exhibitions.