Is tourism destination image uniformly effective in heritage spaces? Spatially paradoxical effects of user-generated content and urban form on tourists’ online preference expressions in Quanzhou Old City, China

Standardized tourism offerings erode the distinctiveness of heritage cities. Yet tourism destination image (TDI) is still widely modeled as spatially uniform across a city, obscuring spatially uneven associations. This study examines how micro-scale spatial morphology relates to TDI-related features derived from user-generated content (UGC) and tourists’ online preference expressions, including photographic check-in preference (PCP) and sentiment expression preference (SEP). Using the Quanzhou World Heritage Site as a case study, we integrate three-dimensional morphology indicators with multimodal UGC to extract visual attributes, thematic features, and expressed sentiments. Local modeling with Multiscale Geographically Weighted Regression (MGWR) captures spatially varying associations, while explainable machine learning with eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanations (SHAP) recovers interpretable relationships. Results reveal that (1) Complex and nonlinear associations exist among spatial morphology, UGC-derived TDI-related features, and online preference expressions. (2) Landmark building imagery density is strongly associated with PCP, and the positive association remains when image-proportion indicators are used. (3) Public space imagery follows a more intricate nonlinear pattern: its marginal contribution to SEP declines over some ranges of the predictor. Exploratory interaction analysis reveals that PCP and SEP follow distinct joint predictive structures. (4) These associations vary substantially across space, indicating that “one-size-fits-all” branding and product provision may misallocate managerial effort. The study advances a spatially explicit understanding of how UGC-derived TDI-related features operate across heritage spaces and offers actionable guidance for destination management, including range-sensitive landmark storytelling, diversification of public-space features to avoid genericity, and context-specific precision interventions that enhance vitality without eroding spatial authenticity.