This study addresses the high degree of subjectivity and the supply–demand imbalance in wetland park science communication. Taking the Nanjing Longpao Yangtze River Provincial Wetland Park as a case study, it develops a data-driven, sustainability-oriented design framework integrating NVivo (qualitative data analysis software), the Analytic Hierarchy Process (AHP), the KANO model, Artificial Intelligence Generated Content (AIGC), and Fuzzy Comprehensive Evaluation (FCE) into a closed-loop methodology. A wetland database was constructed through field research and a literature review. User interviews (N = 73) were analysed using NVivo to distill core themes. AHP, based on 15 expert pairwise comparisons, identified Distinctive Plants (global weight 0.314) and Rare Bird Species (0.275) as priority design elements. The KANO model classified 20 science topics into Must-be (N = 5), One-dimensional (N = 2), and Attractive (N = 8) attributes. AIGC was then employed to generate a three-module promotional brochure (wetland flora, avifauna, and culture). FCE with 60 valid respondents confirmed all three modules achieved an ‘Excellent’ rating, validating the framework’s effectiveness. This study provides a replicable, sustainable design paradigm for wetland science communication.

China’s national parks must protect ecological integrity while enabling appropriate public access, yet recreation suitability and ecological resilience potential are commonly assessed separately. Taking Giant Panda National Park as a case study, this research develops a contemporaneous 2023 spatial assessment that integrates an operational Ecosystem Resilience Index (ERI), a MaxEnt-derived Outdoor Recreation Suitability Index (ORSI), bivariate spatial association, and protection-first management zoning. ERI represents relative ecosystem resilience potential through resistance-related condition, adaptability-related condition, and recovery-related potential. ORSI was estimated from 102 spatially thinned recreation occurrences and ten predictors. Two geographically separated holdouts produced a mean test AUC of 0.960 and mean maximum test TSS of 0.907. ERI and ORSI exhibited a positive global bivariate association (Moran’s I = 0.240, permutation p = 0.001), while local LISA identified neighbourhood-scale departures from this park-wide tendency. Alternative component weights retained strong agreement with the equal-weight ERI (Spearman ρ = 0.972–0.986). Exact park-boundary intersections assigned 6.71% of the analyzed area to potential conflict, 10.64% to strict management, and 13.13% to low-impact recreation assessment. This protection-first framework reveals where park-wide alignment masks local management mismatches and provides an updateable basis for site verification and monitoring.