Numerous studies have examined the impact of the built environment on mental health, yet there remains an underexplored area concerning how microenvironments within educational buildings affect students’ mental well-being from a physical environment standpoint. This paper fills this gap by utilizing data from 440 valid questionnaires to develop regression models that assess students’ perceptions of physical environment factors in college teaching buildings and their impact on anxiety likelihood. This study examined the physical environment of the teaching building’s interior, courtyard, and semi-outdoor areas. Findings indicate that students’ perceptions of specific physical environment factors—such as classroom ventilation (p < 0.01, OR = 0.330), lighting (p < 0.01, OR = 0.444), noise conditions (p < 0.01, OR = 0.415), courtyard thermal comfort (p < 0.01, OR = 0.504), and the views from semi-outdoor areas (p < 0.01, OR = 2.779)—significantly influence the likelihood of experiencing anxiety. Optimal physical conditions are linked to reduced student anxiety. The suitability of the physical environment of teaching buildings is interrelated, and it is urgently necessary to address issues related to unsuitable lighting in window areas of classrooms, as well as problems with ventilation, lighting, and noise caused by the corridor layout within teaching buildings. These insights are crucial for the design and renovation of academic buildings to enhance students’ mental well-being.

How heritage tourism scenes evolve alongside digital mediation and how different forms of mediation are represented in visitor accounts remain insufficiently understood. This study examines the Liangzhu heritage site, centred on the Archaeological Ruins of Liangzhu City, using 7172 visitor-generated online records collected from six major Chinese social media and travel platforms between 2015 and July 2025. A stage-based comparison contrasts the natural tourism stage (2015–2020) with the digitally mediated tourism stage (2020–2025), with records assigned to the corresponding stage-specific corpora based on temporal scope and record-level content eligibility. Word-frequency analysis, grounded-theory coding, social-semantic network analysis, and exploratory lexicon-based polarity analysis are combined to trace the stage-related evolution of the heritage tourism scene as represented in online visitor discourse. The results show that discourse in the digitally mediated tourism stage reflected a reconfiguration, rather than a replacement, of the represented heritage tourism scene: physical heritage remained the experiential anchor, while interpretation, interaction, and reconstructed experience became more prominent. AR-related discourse was associated mainly with guidance, recognition, and site-based interpretation, whereas VR-related discourse was associated with immersion, reconstructed scenes, and presence. Lexicon-derived polarity patterns were uneven: facilities, access, and management showed higher positive-polarity shares, while history, landscape, and overall visitor experience showed more mixed label distributions; some stage contrasts in positive-polarity shares were sensitive to platform composition. The findings extend the application of Cultural Scene Theory to stage-based analysis of digitally mediated heritage tourism by showing how cultural-scene elements are represented and connected in platform-mediated visitor discourse. They also offer evidence-informed implications for sustainable heritage interpretation and integrated visitor management, particularly regarding coherence among physical evidence, digital interpretation, visitor access, and service provision.