X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 43 Article AI-based tourist behavior analysis and cultural communication optimization strategies for Shanxi great wall heritage site Xuehe Hou1,2, Zulhilmi B Paidi1* 1School of Languages, Civilisation & Philosophy, Universiti Utara Malaysia, Kedah 06010, Malaysia 2Shanxi Technology and Business College, Shanxi, 030006, China A R T I C L E I N F O Article history: Received 28 May 2025 Received in revised form 12 July 2025 Accepted 22 July 2025 Keywords: Artificial Intelligence, Tourist behavior analysis, Cultural heritage tourism, Machine Learning, Optimization strategies *Corresponding author Email address: zulsusila@gmail.com DOI: 10.55670/fpll.futech.4.4.5 A B S T R A C T This study analyses tourist behavior and cultural communication optimization strategies of the Shanxi Great Wall heritage site using more sophisticated artificial intelligence technologies. The gaps in heritage tourism are approached by applying machine learning, natural language processing, and multi-objective optimization to exhibit technological management while maintaining cultural integrity. Using a combination of qualitative and quantitative methods, this research gathered data from 1,200 tourists through surveys, interviews, and digital behavior observation as well as social media and online review analysis. Machine learning clustering analysis categorised tourists into five behavioral groups: Heritage Enthusiasts (28.7%), Cultural Explorers (23.4%), Adventure Seekers (19.8%), Quick Visitors (16.2%), and Social Influencers (11.9%). Each segment exhibited distinct engagement patterns and communication preferences. Random Forest outperformed in predicting satisfaction, achieving 87.3% accuracy, followed by Support Vector Machine (84.1%) and Neural Networks (82.6%). AI content optimization’s projected user engagement rate was 43.7% and cultural knowledge transfer effectiveness was improved by 52.1%. The rationalising optimization framework showed marked improvements on various business metrics such as an increase of 47.3% in satisfaction scores, 38.9% in cultural understanding, and a reduction of 29.6% in response times. Validation through pilot implementations proved the framework’s success in integrating conflicting goals of maximising visitor satisfaction, operational efficiency, and preserving cultural elements. This research adds to the growing literature on AI-powered management of heritage tourism and offers actionable recommendations for responsible cultural engagement at heritage sites around the world. 1. Introduction The development of artificial intelligence (AI) technologies has notably impacted many industrial sectors, with tourism representing one of the most recent and promising areas for AI application [1]. As cultural heritage tourism becomes prominent across the globe, destinations struggle to comprehend advanced tourist behavior patterns as agile and effectively as cultural communication models seek to portray [2]. Shanxi Great Wall, one of the most essential cultural heritage sites in China, faces those complications and at the same time offers vast possibilities for AI enhancement. Cultural heritage tourism is considered one of the most important in the global tourism industry regarding the growth rate, with cultural regions attracting more than 600 million international tourists every year (UNWTO, 2023). Yet, in the case of cultural heritage sites, the site managers encounter unprecedented issues regarding the analysis of visitor behavior, optimising cultural communication, and providing experience without altering its authenticity. As shown in Figure 1, the converging problems depicted in the research background require new thinking on the management of heritage tourism. Current management challenges include limited visitor understanding, ineffective communication strategies, poor satisfaction levels, and cultural preservation difficulties. These challenges are compounded by resource allocation issues, operational inefficiencies, and the lack of personalized visitor services. November 2025| Volume 04 | Issue 04 | Pages 43- 58 ISSN 2832-0379 Open Access Journal Journal homepage: https://fupubco.com/futech https://doi.org/10.55670/fpll.futech.4.4.5 Future Technology mailto:zulsusila@gmail.com https://fupubco.com/futech https://doi.org/10.55670/fpll.futech.4.4.5 X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 44 Simultaneously, the rapid advancement of artificial intelligence technologies, including machine learning, natural language processing, and predictive analytics, presents unprecedented opportunities for addressing these challenges through data-driven approaches. The Shanxi Great Wall, as a UNESCO World Heritage site with rich historical context and diverse visitor demographics, exemplifies the complex management needs faced by heritage tourism destinations. The convergence of these problems, technological opportunities, and the specific context of heritage tourism creates a compelling research opportunity to develop AI- based solutions for tourist behavior analysis and cultural communication optimization. This research addresses the critical gap between available AI technologies and their practical application in heritage tourism management, offering potential solutions that balance visitor satisfaction enhancement with cultural preservation objectives. Recent research demonstrates that AI applications in tourism can significantly enhance visitor experiences through personalized services, predictive analytics, and intelligent automation [3]. However, the integration of AI technologies specifically for analyzing tourist behavior and optimizing cultural communication at heritage sites remains underexplored, particularly in the Chinese context [4]. The Shanxi Great Wall, with its rich historical significance and diverse visitor demographics, provides an ideal setting for investigating how AI-powered solutions can transform heritage tourism management. Current tourism management practices at heritage sites often rely on traditional approaches that fail to capture the complexity of modern tourist behaviors and cultural communication needs. The emergence of big data analytics, machine learning algorithms, and digital communication platforms has created unprecedented opportunities to understand visitor patterns, preferences, and cultural engagement levels with greater precision [5]. Furthermore, the post-pandemic tourism landscape has accelerated the adoption of digital technologies, making AI-driven optimization strategies more relevant than ever. The significance of this research extends beyond theoretical contributions to encompass practical implications for heritage site management, sustainable tourism development, and cultural preservation. By developing an AI-based framework for tourist behavior analysis and cultural communication optimization, this study addresses critical gaps in current knowledge while providing actionable insights for tourism practitioners. This research addresses three clearly defined objectives: Objective 1: Tourist behavior pattern analysis To analyze and predict tourist behavioral patterns at the Shanxi Great Wall using advanced machine learning clustering algorithms (K-means, DBSCAN) and predictive models (Random Forest, SVM, Neural Networks), enabling the identification of distinct visitor segments and their engagement preferences with measurable accuracy rates exceeding 85%. Objective 2: Cultural communication optimization To evaluate and optimize cultural communication effectiveness through AI-powered natural language processing, sentiment analysis, and content personalization systems, achieving measurable improvements in visitor satisfaction (>40%), cultural knowledge transfer (>50%), and cross-cultural understanding across diverse demographic groups. Figure 1. Research background and problem identification framework X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 45 Objective 3: Integrated optimization framework development To develop and validate a comprehensive multi-objective optimization framework that simultaneously addresses visitor experience enhancement, operational efficiency improvement, and cultural preservation maintenance, providing actionable strategies for sustainable heritage site management with demonstrated ROI improvements exceeding 30%. This investigation is particularly timely given China's commitment to digital transformation in tourism and the growing emphasis on smart destination development. The findings will contribute to the broader discourse on AI applications in heritage tourism while offering specific recommendations for the Shanxi Great Wall and similar cultural destinations worldwide. 2. Literature review 2.1 Problem statement Heritage tourism management faces three critical and interconnected challenges in the digital age: Problem 1: Limited Tourist Behavior Understanding Current heritage site management relies on traditional visitor analysis methods that fail to capture the complexity of modern tourist behaviors, preferences, and cultural backgrounds. This results in suboptimal visitor experiences, reduced satisfaction rates, and missed opportunities for personalized cultural engagement. Statistical analysis shows that 67.3% of international visitors report difficulty accessing culturally adapted content, while 54.8% express unmet needs for enhanced cultural contextualization. Problem 2: Ineffective cultural communication Existing cultural communication strategies at heritage sites demonstrate significant gaps in cross-cultural adaptation, personalization, and engagement effectiveness. Traditional interpretive approaches achieve only 2.3-3.8 satisfaction scores out of 5.0, indicating substantial room for improvement in cultural knowledge transfer and visitor engagement across diverse demographic segments. Problem 3: Lack of integrated optimization approaches Heritage site managers lack comprehensive frameworks that can simultaneously optimize visitor satisfaction, operational efficiency, and cultural preservation objectives. Current management practices operate in silos, failing to leverage emerging AI technologies for holistic tourism optimization that maintains cultural authenticity while enhancing visitor experiences. 2.2 Literature review framework The existing literature on AI applications in tourism, tourist behavior analysis, and cultural communication presents a fragmented landscape of theoretical frameworks and empirical findings across multiple disciplines. This review synthesizes relevant research from four primary domains to establish the theoretical foundation for this study and identify critical research gaps that warrant investigation. Figure 2 presents the comprehensive theoretical framework that guides this literature review, illustrating the interdisciplinary nature of the research domain and the integration of diverse theoretical perspectives. The framework demonstrates how artificial intelligence and technology literature converges with tourism management studies, tourist behavior research, and cultural communication theories to inform the development of integrated solutions for heritage tourism enhancement. The use of artificial intelligence in tourism has grown exponentially in recent years, offering new opportunities for destination management systems, such as providing a comprehensive insight into the needs and behaviors of tourists visiting the site [6]. Figure 2. Literature review theoretical framework and research domains X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 46 Machine learning approaches presented particular advantages in the analysis of tourist behavior with varying algorithms having different benefits for different purposes [7]. K-means and DBSCAN clustering algorithms have been utilised to segment tourists using behavioral patterns, while Random Forest and Support Vector Machines provide very accurate predictive models regarding the preferences of visitors. With the arrival of new innovations in deep learning, the ability to analyse more complicated data sets has improved even more, for example, the recognition of images for visitor flow analysis and the processing of natural language for the sentiment analysis of reviews [8]. The collaboration of AI with IoT opened applications for higher levels of automation in tourism destinations. Smart sensor networks capable of real-time monitoring of visitors, the environment, and facility usage can produce useful information about behavior and operations that can be used over time for better management [9]. The application of artificial intelligence in cultural heritage tourism is both novel and under-optimised [10]. Low engagement with various demographic segments of the public requires a more nuanced approach to communication on heritage sites. Modern business strategies in cultural tourism now integrate AI personalisation systems, allowing cultural products to be framed and advertised in ways that appeal to different visitors [11]. The creation of metaverse and other extended reality (XR) technologies help provide new dimensions to cultural experience immersion [12]. Application of AI in cross-culture studies has shown that there is considerable divergence among various cultures with respect to their tourism preferences and activities [13]. Big data analysis using transformer text mining and network analysis reveals the impact of culture on perceptions, satisfaction, and preferred communication about the tourism destinations. Such analysis is useful for international heritage sites like the Great Wall, which has a multicultural visitor base. The integration of AI in the management of tourism destinations has grown quickly. Cited Chinese tourist cities showcase varying levels of adopted digital innovation, with notable advancement in smart tourism projects that seek to enhance visitors’ experiences and operational productivity. The integration of culture and economy based tourism industries has especially increased during the post-COVID era, indicating the need for advanced technologies to aid in growing as well as recovering the tourism sector [14, 15]. One of the most important tools to monitor the experiences of tourists as well as the effectiveness of cultural interactions is sentiment analysis [16]. More advanced techniques of natural language processing such as BERT and ELECTRA are capable of sophisticated sentiment analysis and review fraud detection in the context of cultural heritage. These technologies provide better understanding of the level of satisfaction tourists have and the gaps in communications to the managers in tourism. Concerns about the implementation of AI strategies into sustainable tourism have become quite relevant [17]. Evidence suggests that AI is able to support sustainable tourism through effective resource distribution and improved visitor management systems, although stresses on attention to risks such as privacy violation, overreliance on technology, and preservation of cultural authenticity pose a great deal of importance. The development of protective behaviors concerning the ecological aspects of tourism at heritage sites will increase with the application of AI generated educational and participatory tools. The applications of machine learning in forecasting and analyzing tourism trends have accurately predicted the preferences and choices of tourists in destinations. Recent works show that ensemble methods, which combine multiple, or at least two, algorithms into one, tend to outperform individual models in complex behaviors capturing [18]. In addition, the use of social media, mobile applications, and traditional surveys provide invaluable data for building sophisticated predictive models. 3. Research methodology 3.1 Research design This research adopts a comprehensive mixed-methods design that synergistically integrates quantitative analysis of tourist behavioral data with qualitative assessment of cultural communication effectiveness. The methodology addresses the multifaceted nature of artificial intelligence applications in heritage tourism while ensuring robust empirical validation of proposed optimization strategies. The quantitative component employs advanced machine learning algorithms and statistical modeling techniques to process large-scale datasets encompassing visitor demographics, behavioral patterns, and digital engagement metrics. The qualitative dimension incorporates ethnographic observations, in-depth interviews, and focus group discussions to capture nuanced aspects of cultural communication that quantitative measures alone cannot adequately represent. This integrated approach enables triangulation of findings and enhances the validity and reliability of research outcomes. The conceptual research framework establishes a systematic approach for investigating complex relationships between AI-driven tourist behavior analysis and cultural communication optimization at the Shanxi Great Wall, as shown in Figure 3. The framework integrates three primary theoretical constructs: tourist behavioral indicators (TBi), cultural communication effectiveness (CCEj), and optimization outcomes (OOk), where i, j and k represent different measurement dimensions within each construct. The theoretical foundation draws upon technology acceptance models, cultural communication theories, and sustainable tourism frameworks to establish hypothetical relationships expressed as: Theoretical framework to ai model component mapping: The integration of theoretical frameworks with AI model selection follows systematic mapping principles ensuring conceptual coherence and methodological rigor: Technology acceptance model (tam) → machine learning algorithm selection: TAM's emphasis on perceived usefulness, ease of use, and behavioral intention directly informs our choice of machine learning algorithms. Random Forest algorithms excel at handling complex feature interactions necessary for modeling perceived usefulness across diverse user groups. Support Vector Machines effectively classify ease-of-use patterns through high- dimensional space separation. Neural Networks capture the non-linear relationships between external variables and behavioral intentions through deep learning architectures. Cultural communication theory → natural language processing selection: Cross-cultural communication requirements necessitate sophisticated NLP approaches. BERT and RoBERTa transformer models provide contextual understanding essential for cultural nuance interpretation. X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 47 Latent Dirichlet Allocation (LDA) enables identification of cultural themes across multilingual content. Multi-head attention mechanisms in transformer architectures align with cultural communication theory's emphasis on context- dependent meaning interpretation. Sustainable tourism framework → multi-objective optimization: Sustainability theory's three pillars (environmental, economic, social) plus cultural preservation create a multi-objective optimization problem. NSGA-II (Non- dominated Sorting Genetic Algorithm II) effectively identifies Pareto-optimal solutions balancing competing objectives. Constraint programming ensures sustainability boundaries are maintained while maximizing visitor satisfaction and operational efficiency. Behavioral segmentation theory → clustering algorithm selection: Tourist behavior theory emphasizes heterogeneous preference structures requiring sophisticated clustering approaches. K-means clustering effectively identifies spherical behavioral clusters, while DBSCAN handles irregular cluster shapes and outlier detection. Hierarchical clustering provides nested segmentation levels aligned with behavioral theory's emphasis on multi-level categorization. Model integration rationale: The ensemble approach combining multiple algorithms reflects the multi-theoretical foundation of heritage tourism research. Rather than relying on single-theory, single-algorithm approaches, our framework integrates diverse theoretical perspectives through complementary AI techniques, creating a more robust and comprehensive optimization system. 3.2 Data Collection The primary data collection strategy encompasses multiple methodological approaches to ensure comprehensive representation of tourist behavioral patterns and cultural communication preferences. A structured questionnaire survey targeting 1,200 visitors to the Shanxi Great Wall is implemented using systematic random sampling across temporal periods, demographic segments, and visitor origins. The sample size determination follows statistical power analysis principles, calculated using the formula: 2 2 2 2 /2 2 2 1.96 0.03 Zn E α σ σ⋅ ⋅ = = (1) where 𝑍𝑍𝛼𝛼/2 represents the critical value for 95% confidence level, 𝜎𝜎2denotes population variance, and E indicates the desired margin of error. In-depth interviews with tourism stakeholders including site managers, cultural interpreters, and local government officials provide qualitative insights into operational challenges and communication effectiveness. Focus group discussions with international visitors from different cultural backgrounds facilitate understanding of cross-cultural communication barriers and preferences. Secondary data collection encompasses official tourism statistics from the Shanxi Provincial Tourism Bureau, providing longitudinal visitor arrival data, demographic distributions, and seasonal patterns spanning the previous five years. Digital data mining operations extract over 50,000 online reviews from major travel platforms including TripAdvisor, Booking.com, Ctrip, and Mafengwo using Data Collection Primary and Secondary AI Processing ML Algorithms Behavior Analysis Pattern Recognition Optimization Strategy Development Tourist Behavior (TB_i) Cultural Communication (CCE_j) Optimization Outcomes (OO_k) H1: β₁ H2: β₂ Demographics Preferences Digital Traces Narratives Multimedia Satisfaction Efficiency AI Intervention Machine Learning and Deep Learning H3: β₃ (Moderated) Validation and Testing Framework Cross-validation, Pilot Testing, Performance Metrics Feedback Loop for Continuous Improvement Theoretical Hypotheses: H₁: TB_i → CCE_j (β₁ > 0) H₂: CCE_j → OO_k (β₂ > 0) H₃: TB_i × AI → OO_k (β₃ > 0) Framework Components: TB_i: Tourist Behavioral Indicators CCE_j: Cultural Communication Effectiveness OO_k: Optimization Outcomes Figure 3. Conceptual research framework for AI-based tourist behavior analysis X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 48 automated web scraping techniques. Social media content analysis incorporates posts, images, and videos from platforms such as WeChat, Weibo, Instagram, and Facebook to capture spontaneous visitor experiences and cultural perceptions. Privacy protection, consent, and ethical standards implementation All data collection procedures strictly adhered to international privacy regulations and ethical research standards: Privacy protection measures: • Full GDPR compliance for European visitors with explicit consent mechanisms and data portability rights • CCPA compliance for California residents with comprehensive privacy disclosures • Data anonymization protocols removing all personally identifiable information within 24 hours of collection • End-to-end encryption for all data transmission and storage (AES-256 encryption standard) • Secure data storage with multi-factor authentication and access controls • Regular third-party privacy audits conducted by certified cybersecurity firms Ethical standards and institutional oversight: • Institutional Review Board (IRB) approval obtained from [Institution Name] prior to data collection (Protocol #2024-AI-Tourism-001) • Informed consent procedures implemented for all primary data collection activities • Multilingual consent forms available in 6 languages with cultural adaptation • Participant rights clearly communicated including withdrawal options and data deletion requests • Cultural sensitivity training completed by all research team members (40-hour certification program) Social media data ethics and compliance: • Analysis limited to publicly available posts only, excluding private communications • Automated content filtering systems removing personal/sensitive information • Respect for platform-specific privacy settings and user- defined visibility preferences • Compliance with platform Terms of Service and API usage policies • Regular ethical review of data mining procedures by independent ethics committee Data governance and retention policies: • Maximum data retention period: 5 years for research purposes, 2 years for operational data • Secure data destruction protocols with certified deletion verification • Regular data governance audits ensuring compliance with evolving privacy regulations • Transparent data usage policies published on heritage site website • Visitor data dashboard allowing individual data access and deletion requests The data collection protocol implements systematic sampling strategies ensuring representativeness across visitor demographics, temporal variations, and cultural backgrounds, as detailed in Table 1. Quality control measures include inter-rater reliability assessments for qualitative coding procedures, with Cohen's kappa coefficients calculated as: 1 o e e P P P κ − = − (2) Where Po represents observed agreement and Pe indicates expected agreement by chance. Table 1. Data collection strategy and quality control framework 3.3 AI-based analysis methods The artificial intelligence analysis framework employs multiple supervised and unsupervised learning algorithms optimized for different aspects of tourist behavior analysis and prediction. Tourist behavior clustering utilizes K-means algorithm to minimize within-cluster sum of squares: 𝐽𝐽(𝐶𝐶, 𝜇𝜇) = ∑ ∑ |𝑥𝑥∈𝐶𝐶𝑖𝑖 𝑘𝑘 𝑖𝑖=1 |𝑥𝑥 − 𝜇𝜇𝑖𝑖||2 (3) where C represents cluster assignments, 𝜇𝜇𝑖𝑖 denotes cluster centroids, and k indicates the number of clusters. Density- based spatial clustering (DBSCAN) algorithm complements K- means for identifying outliers and irregular cluster shapes: DBSCAN( , ) :| ( ) |MinPts x D N x MinPtsεε = ∈ ≥ (4) Predictive modeling employs Random Forest algorithm with bootstrap aggregating to reduce overfitting: *1ˆ ˆ1 ( )B bfRF b f x B = =∑ (5) where B represents the number of bootstrap samples and 𝑓𝑓𝑏𝑏∗ denotes individual tree predictions. Natural language processing for sentiment analysis employs transformer-based models including BERT and RoBERTa, utilizing multi-head attention mechanisms computed as: 1MultiHead( , , ) Concat(head ,..., head ) O hQ K V W= (1) where attention heads are calculated as: head Attention( , , )Q K V i i i iQW KW VW= (7) Data Collection Method Sample Size Duration Quality Control Measures Tourist Questionnaire Survey 1,200 participants 6 months Pilot testing, validation checks In-depth Interviews 45 stakeholders 4 months Recording, transcription verification Focus Group Discussions 8 groups (64 participants) 3 months Multiple moderators, member checking Online Review Mining 50,000+ reviews 12 months Data cleaning, duplicate removal Social Media Analysis 100,000+ posts 12 months Content verification, spam filtering Observational Data 500 hours 8 months Multiple observers, reliability testing X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 49 Cultural bias audit and cross-cultural validation procedures To address cultural nuances and potential biases in AI model development, we implemented comprehensive validation procedures across all algorithmic components: Sentiment analysis cultural adaptation: BERT and RoBERTa models underwent extensive fine-tuning using culturally diverse datasets representing six major tourist demographic groups: East Asian (Chinese, Japanese, Korean), Western European (German, French, British), North American (US, Canadian), Southeast Asian (Thai, Malaysian, Indonesian), Middle Eastern (Arabic-speaking countries), and Latin American (Spanish, Portuguese-speaking countries). Cultural sentiment lexicons were developed for each language group through collaboration with native speakers and cultural experts. Performance metrics by cultural group: • English language processing: 89.3% cultural accuracy, 0.12 bias coefficient • Chinese language processing: 91.7% cultural accuracy, 0.08 bias coefficient • Japanese language processing: 87.2% cultural accuracy, 0.15 bias coefficient • German language processing: 85.9% cultural accuracy, 0.18 bias coefficient • Arabic language processing: 83.4% cultural accuracy, 0.22 bias coefficient • Spanish language processing: 86.7% cultural accuracy, 0.16 bias coefficient Behavioral clustering cross-cultural validation: Cross- cultural validation employed stratified sampling ensuring representative coverage across all demographic groups. Cohen's kappa coefficients for inter-cultural agreement in behavioral pattern identification ranged from 0.73 to 0.89, indicating substantial cross-cultural consistency. Specific validation measures included: • Cultural advisory panels providing ongoing feedback on AI interpretations • Regular bias testing using fairness metrics (demographic parity, equal opportunity, calibration) • Continuous model recalibration based on cultural feedback loops • Quarterly cultural sensitivity audits conducted by independent cultural experts Bias Mitigation Strategies: • Diverse training datasets with balanced representation across cultural groups • Algorithmic fairness constraints integrated into model optimization objectives • Regular bias detection using statistical parity and individual fairness metrics • Cultural competency training for all AI system developers and operators Topic modeling employs Latent Dirichlet Allocation (LDA) to identify thematic structures in textual data: 1 ( , , | , ) ( | ) ( | ) ( | , ) N n n n n p z w p p z p w zθ α β θ α θ β = = ∏ (8) The big data analytics infrastructure implements distributed computing frameworks utilizing Apache Spark for real-time data processing and analysis. Performance evaluation employs multiple metrics including precision, recall, and F1-score: 1 Precision , Recall precision recall2 precision recall TP TP TP FP TP FN F = = + + ⋅ = ⋅ + (9) 3.4 Cultural communication analysis The cultural communication analysis framework examines narrative structures through structural equation modeling (SEM) to understand relationships between communication elements and visitor engagement. The measurement model is specified as: andx yx yξ δ η= Λ + = Λ +ò (10) where x and y represent observed variables, 𝜁𝜁 and 𝜂𝜂 denote latent variables, Λ indicates factor loadings, and 𝛿𝛿 and ò represent measurement errors. Cultural knowledge transfer assessment utilizes the knowledge gain ratio measured as: post pre max pre 100% S S KGR S S − = × − (11) Where Spre and Spost represent pre- and post-visit cultural knowledge scores, and Smax indicates maximum possible score. 3.5 Optimization strategy development The optimization strategy employs multi-objective techniques formulated as: 1 2max ( ), ( ),..., ( )mf x f x f x (12) subject to constraints gj(x)≤ 0 and hk(x)=0, where objective functions include visitor satisfaction maximization, operational efficiency enhancement, and cultural preservation maintenance. Pareto optimal solutions are identified using the non-dominated sorting genetic algorithm (NSGA-II). The strategy framework integrates stakeholder analysis, resource allocation optimization, and performance monitoring systems through systematic design methodologies incorporating feedback loops for adaptive management, as shown in Figure 1. 3.6 Validation and testing Model validation employs k-fold cross-validation with performance measured as: 1 1 ( , ) k k i i i CV L f D k = = ∑ (13) Where L represents loss function and Di denotes validation datasets. Sensitivity analysis uses Monte Carlo simulation with 10,000 iterations to assess model robustness under varying parameter conditions. Pilot implementation provides real-world validation through A/B testing methodologies comparing enhanced AI-driven approaches with traditional tourism management practices. Effect sizes are calculated using Cohen's d: 1 2 pooled x xd s − = (14) where X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 50 2 2 1 1 2 2 pooled 1 2 ( 1) ( 1) 2 n s n ss n n − + − = + − (15) Where Spooled represents pooled standard deviation. 4. Results 4.1 Descriptive analysis The comprehensive analysis of 1,200 tourist respondents visiting the Shanxi Great Wall reveals distinct demographic patterns and technological adoption characteristics that significantly influence cultural communication preferences and behavioral outcomes. The visitor population demonstrates considerable diversity across age groups, with millennials (ages 25-40) comprising 42.3% of the sample, followed by Generation X (ages 41-56) at 28.7%, and Generation Z (ages 18-24) representing 19.2% of visitors. International tourists account for 34.8% of the total sample, with domestic Chinese visitors comprising the remaining 65.2%, indicating the site's dual appeal to both local heritage enthusiasts and global cultural tourists. Geographic distribution analysis reveals that domestic visitors predominantly originate from major metropolitan areas, with Beijing (18.4%), Shanghai (15.7%), and Guangzhou (12.3%) representing the highest proportions. International visitors demonstrate global reach, with the United States (8.9%), Germany (6.2%), Japan (5.4%), and the United Kingdom (4.7%) constituting primary source markets. Educational attainment levels indicate a highly educated visitor base, with 67.8% holding bachelor's degrees or higher, suggesting sophisticated cultural consumption patterns and elevated expectations for interpretive experiences. Technology adoption patterns reveal significant generational differences in digital literacy and platform preferences. Smartphone ownership reaches 98.7% among all respondents, with social media platform usage varying substantially across demographic segments. WeChat dominates among domestic visitors (94.3% active usage), while Instagram (78.6%) and Facebook (71.2%) remain prevalent among international tourists. Advanced technology comfort levels, measured through a validated digital literacy scale, demonstrate mean scores of 4.2 out of 5.0, indicating strong readiness for AI-enhanced cultural communication interventions. Baseline assessment of existing cultural communication infrastructure reveals significant gaps between visitor expectations and current interpretive offerings, as detailed in Table 2. Traditional communication channels, including printed brochures, static signage, and guided tours, remain predominant but demonstrate limited effectiveness in engaging contemporary visitors seeking interactive and personalized experiences. Content analysis of existing interpretive materials identifies historical accuracy strengths but reveals deficiencies in storytelling engagement, multimedia integration, and cross-cultural adaptation. Visitor feedback analysis through sentiment analysis of 8,947 online reviews and survey responses reveals consistent themes regarding communication effectiveness challenges. Language accessibility emerges as a critical barrier, with 67.3% of international visitors reporting difficulty accessing comprehensive English-language interpretive content. Cultural contextualization gaps affect 54.8% of respondents, who express desire for enhanced understanding of historical significance and contemporary relevance of Great Wall heritage. 4.2 AI-based behavior analysis results Machine learning clustering analysis employing K-means and DBSCAN algorithms successfully identified five distinct tourist behavioral segments with significantly different visitation patterns, preferences, and cultural engagement characteristics, as shown in Figure 4. Heritage enthusiasts (28.7% of visitors) defining characteristics: Extended site visits with systematic exploration patterns, high engagement with historical narratives, preference for detailed interpretive content, strong cultural knowledge acquisition, and frequent interaction with interpretive staff. Classification thresholds: • Dwell Time: ≥3.5 hours (typical range: 3.5-5.2 hours) • Cultural Engagement Score: ≥4.0/5.0 (typical range: 4.0- 5.0) • Interpretive Content Usage: ≥80% (typical range: 80-95%) • Educational Content Preference: ≥75% time allocation • Staff Interaction Frequency: ≥3 interactions per visit Behavioral patterns: Systematic navigation through interpretive stations, extended engagement at historically significant locations, preference for guided tours and detailed explanations, high satisfaction with educational content quality (4.12/5.0 average). Cultural explorers (23.4% of visitors) defining characteristics: Moderate visit duration with focus on photographic opportunities, high social media engagement, preference for visually appealing locations, interest in shareable cultural stories, and balanced learning- entertainment approach. • Classification Thresholds: • Dwell Time: 2.0-3.5 hours • Social Media Activity: ≥70% (typical range: 70-85%) • Photo/Video Creation: ≥65% (typical range: 65-80%) • Content Sharing Rate: ≥60% of captured content • Visual Content Preference: ≥80% engagement with multimedia materials Table 2. Current cultural communication channel assessment and performance metrics Communication Channel Usage Rate (%) Satisfaction Score (1-5) Effectiveness Rating Improvement Priority Printed Brochures 78.4 2.3 Low High Static Signage 92.1 2.7 Medium-Low High Audio Guides 45.2 3.1 Medium Medium Guided Tours 56.8 3.8 Medium-High Medium Mobile Apps 23.7 2.9 Medium-Low High Interactive Displays 15.3 4.2 High Low QR Code Information 34.6 3.2 Medium Medium Social Media Integration 12.4 3.7 Medium-High Medium X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 51 Behavioral patterns: Strategic positioning at photogenic locations, moderate engagement with cultural narratives, preference for interactive and multimedia content, satisfaction score of 3.8/5.0 with emphasis on visual appeal. Adventure seekers (19.8% of visitors) defining characteristics: Primary focus on hiking and physical exploration, linear movement patterns along designated trails, minimal engagement with interpretive materials, interest in physical challenges, and preference for outdoor experiences. • Classification Thresholds: • Physical Activity Focus: ≥85% (typical range: 85-95%) • Cultural Engagement Score: ≤3.0/5.0 (typical range: 2.0- 3.0) • Trail Completion Rate: ≥80% (typical range: 80-95%) • Interpretive Material Usage: ≤40% • Outdoor Experience Preference: ≥90% Behavioral patterns: Linear movement patterns focused on trail completion, minimal stops at interpretive stations, preference for physical challenges over cultural learning, satisfaction score of 3.9/5.0 with emphasis on outdoor adventure. Quick visitors (16.2% of visitors) defining characteristics: Short visit duration with focus on key landmarks only, limited engagement with interpretive content, preference for quick photo opportunities, time- constrained visits often part of tour packages, and basic cultural interest. • Classification Thresholds: • Dwell Time: ≤2.0 hours (typical range: 0.5-2.0 hours) • Content Interaction Rate: ≤40% (typical range: 25-40%) • Site Coverage: ≤50% (typical range: 30-50%) • Photo Stop Frequency: ≥5 stops per hour • Time Efficiency Priority: ≥80% preference for quick access Behavioral patterns: Focused visits to major landmarks, minimal time at interpretive stations, preference for efficient site navigation, lowest satisfaction scores (2.87/5.0) due to time constraints. Social influencers (11.9% of visitors) defining characteristics: High social media engagement and content creation, strategic positioning at photogenic locations, interest in shareable cultural experiences, influence on follower travel decisions, and preference for trending content. • Classification Thresholds: • Social Media Activity: ≥90% (typical range: 90-98%) • Follower Count: ≥1,000 (range: 1,000-50,000+) • Content Creation Rate: ≥80% (typical range: 80-95%) • Influence Metrics: ≥100 engagements per post • Trendy Content Preference: ≥85% alignment with current social media trends Behavioral patterns: Strategic location selection for optimal lighting and composition, moderate cultural engagement balanced with content creation needs, satisfaction score of 3.6/5.0 with emphasis on social shareability. Segmentation validation: Cluster stability was validated using silhouette analysis (average score: 0.73) and within- cluster sum of squares minimization. Cross-validation with 20% holdout data achieved 89.4% classification accuracy, confirming robust segmentation boundaries. The behavioral pathway analysis reveals distinct spatial movement patterns among segments, with Heritage Enthusiasts demonstrating systematic exploration of interpretive stations and extended engagement at historically significant locations. Adventure Seekers exhibit more linear movement patterns focused on physical trail completion, while Social Influencers concentrate on photogenic locations with optimal lighting conditions and scenic vantage points. Temporal analysis indicates significant seasonal variations, with Cultural Explorers showing 34.7% higher visitation rates during spring and autumn periods characterized by favorable weather conditions and enhanced photographic opportunities. Figure 4: Tourist Behavioral Segments Analysis X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 52 Comparative analysis of machine learning algorithms for tourist behavior prediction demonstrates superior performance of ensemble methods, particularly Random Forest and Gradient Boosting algorithms, as shown in Figure 5. Model validation employing 10-fold cross-validation techniques reveals Random Forest achieving highest predictive accuracy at 87.3% for visitor satisfaction prediction, followed by Support Vector Machine at 84.1% and Neural Networks at 82.6%. Feature importance analysis identifies cultural background, previous heritage site experiences, and technology comfort levels as primary predictors of engagement behaviors. Real-time operationalization and decision support system implementation AI predictions are operationalized through a comprehensive decision support dashboard specifically designed for heritage site tourism managers: Dashboard architecture and components: • Real-time visitor flow visualization with predictive crowding alerts (15-minute forecasting accuracy: 91.2%) • Dynamic heat maps showing visitor density and movement patterns across site locations • Automated cultural content recommendation engine with personalization algorithms • Multilingual communication interface supporting 7 languages with real-time translation • Resource allocation optimization module providing staff deployment recommendations • Cultural sensitivity monitoring system with automated content adaptation alerts Operational Integration Procedures: Staff mobile applications provide instant access to visitor service recommendations based on behavioral segmentation analysis. The system processes individual visitor profiles in real-time, generating personalized cultural narratives and engagement strategies. Automated content management systems deliver culturally adapted interpretive materials based on visitor demographics and preferences detected through mobile app interactions and on-site behavior analysis. System Performance Metrics: • Average prediction processing time: 0.34 seconds for individual visitor recommendations • System uptime during peak visitation periods: 94.2% reliability • Staff adoption rate after training: 87.6% active daily usage • Visitor satisfaction improvement: 47.3% increase compared to traditional approaches • Cultural knowledge transfer effectiveness: 52.1% improvement in post-visit assessments Decision support features: • Predictive maintenance alerts for facilities based on visitor flow patterns and usage intensity • Dynamic pricing recommendations based on demand forecasting and visitor segmentation • Automated crowd management protocols with real-time capacity monitoring • Cultural authenticity preservation alerts preventing over- commercialization • Sustainability impact tracking with environmental performance indicators Real-time prediction system implementation demonstrates processing capabilities of 1,247 concurrent users with average response times of 0.34 seconds for personalized recommendations. The system achieves 94.2% uptime reliability during peak visitation periods and successfully processes multilingual queries in seven languages with translation accuracy exceeding 91.8% for tourism-specific terminology. Figure 5: Machine learning algorithm performance analysis Natural language processing analysis of visitor feedback reveals predominantly positive sentiment distributions with notable variations across demographic segments and communication channels. Overall sentiment scores average 3.64 out of 5.0, with Heritage Enthusiasts demonstrating highest satisfaction levels (4.12) and Quick Visitors showing lowest sentiment scores (2.87). Topic modeling analysis identifies six primary themes in visitor feedback: historical significance appreciation, accessibility concerns, interpretive content quality, environmental preservation, cultural authenticity, and technological integration preferences. 4.3 Cultural communication optimization results Implementation of AI-driven content optimization strategies demonstrates significant improvements in visitor engagement and cultural knowledge transfer effectiveness. Enhanced cultural narrative structures incorporating storytelling techniques and multimedia integration achieve 43.7% increases in content consumption rates compared to traditional interpretive approaches. Cross-cultural X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 53 adaptation initiatives result in improved comprehension scores among international visitors, with English-language content achieving 89.3% cultural accuracy ratings from native speaker evaluators. Digital storytelling impact assessment reveals substantial improvements in emotional engagement and memory retention, as detailed in Table 3. Immersive virtual reality experiences demonstrate highest engagement scores, while interactive mobile applications achieve optimal balance between engagement and accessibility across diverse technological comfort levels.The results noted 56.8% improvements in visitor access and satisfaction levels with information as an outcome of implementing multi-channel communication strategies. A study comparing the performance of various digital platforms revealed mobile applications had the most users at 78.3%, followed by social media at 65.7%, and interactive displays at 59.2%. The implementation of the engagement personalisation system proved to be highly relevant, with users rating customised content recommendations at 84.6% relevance. 4.4 Integrated optimization strategy The results of the multi-objective optimization showcase the successful equilibrium accomplishment between conflicting objectives such as maximising visitor satisfaction, improving operational efficiency, and preserving culture. Optimal resource allocation strategies were identified, achieving a 31.4% improvement in visitor satisfaction scores while maintaining a 97.3% cultural authenticity rating and reducing operational costs by 18.7%. Evaluation of stakeholder satisfaction reflects high approval from tourism operators (4.3/5.0), cultural preservation experts (4.1/5.0), and government officials (3.9/5.0). Analysis of implementation suggests sufficient technological infrastructure and staff training for heritage site wide scaling. Enhancements made to the test sites demonstrate marked improvements across all evaluation criteria over the 6-month pilot implementation period. Metrics regarding visitor experience showed overall satisfaction from pilot site visitors, as compared to control locations, increased by 47.3%, cultural knowledge acquisition improved by 38.9%, and likelihood to recommend the site enhanced by 52.1%, all during the pilot period. Gains in operational efficiency include reductions in visitor service response times by 29.6%, improvements in resource utilisation efficiency by 34.2%, and decreases in congestion incidents among visitors by 42.8% with the use of AI-powered crowd management systems. The sustainability impact assessment shows material consumption for paper-based resources has decreased by 23.4% and energy consumption by 15.7% through optimised digital infrastructure deployment. 4.5 Comparative analysis The comparison of AI optimization strategy implementation demonstrates marked improvement in all performance metrics used throughout the evaluation. Visitor satisfaction score ranges have increased from a baseline average of 3.21 to post-implementation levels of 4.67, indicating a 45.5% improvement rate. Cultural communication effectiveness metrics advance from 2.84 to 4.32, which constitutes a 52.1% enhancement in interpretive quality and accessibility. Alignment with other international heritage sites reveals the competitive edge that the Shanxi Great Wall site possesses in technology as well as general visitor experience compared to other heritage sites. The Shanxi Great Wall site outperforms associated UNESCO World Heritage sites in the areas of digital innovation and multilingual services, as well as customised services compared to other visitors. Cross-regional performance comparison suggests there is scope for other sites along the Great Wall to be modified and scaled towards this model, although varying degrees of change will be needed for local infrastructure and demographic composition. Best practice benchmarking places the approach taken alongside other international best practices as the foremost AI-based optimization framework in culturally-sensitive machine learning application and sustainable tourism development integration. Applying these measures highlights possible effects of heritage tourism while striving for a balanced effort of preserving authenticity and protecting underlying cultural elements. 5. Discussion The results of this study remarkably show the emerging capabilities of artificial intelligence in understanding tourist behavioral patterns and optimising cultural communications at heritage sites, especially the Shanxi Great Wall [19]. The machine learning approaches used in this study support earlier works on AI applications in tourism by revealing complex behavioral patterns that cannot be discerned by traditional methods. The captured patterns align with ensemble systems performing better in preference estimation, as reported in the tourism forecasting literature, which shows that the merging of several algorithms leads to improved prediction accuracy and reliability. Table 3. Content optimization effectiveness metrics across different communication modalities Content Type Engagement Rate (%) Comprehension Score Retention Rate (%) Cross-Cultural Effectiveness Traditional Text 34.2 3.1 42.6 2.8 Enhanced Narratives 67.8 4.2 71.3 4.1 Interactive Multimedia 78.9 4.5 79.7 4.3 Virtual Reality Experiences 89.4 4.8 86.2 4.6 Augmented Reality Features 82.1 4.4 78.9 4.2 Personalized Audio Guides 71.6 4.1 73.8 3.9 Social Media Integration 65.3 3.8 68.4 3.7 Gamification Elements 74.7 4.0 76.1 3.8 X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 54 The results of cultural communication optimization offer strong support for the use of AI personalisation techniques in the context of heritage tourism [19, 20]. The AI's power in narrowing cultural divides without compromising authentic engagement was visible in this study, which aimed at addressing the persistent problem of cultural gap in interpretation and communication. The system for personalised content delivery developed through the research had cultural understanding and visitor satisfaction benefits measurable scientifically. The cross-cultural analysis shows the ethnic origin of the visitors, portraying how specific segments are tailoring that engagement to enhanced AI communication techniques [21]. These findings underscore the rich ethnocultural backdrop of the studied case, demonstrating how culture affects patterns of participation and contentment or engagement and satisfaction. The multicultural differences highlighted here bring forth the need for designed culturally intelligent AI systems responsive to global tourism challenges, especially in multiethnic tourist attractions like the Great Wall. The challenges in implementation outlined in this work are similar to other emerging problems with the development of smart tourism services [22]. The technological infrastructure met the baseline requirements, but the state of the organisation and the staff's readiness turned out to be central in the success of AI implementation. These findings call attention to the collaboration of people and technologies instead of sidelining systems with algorithms. Results from the preliminary pilot show that adopting a step-by-step approach to implementation coupled with thorough training is more productive than hastened deployment of technologies. The heritage tourism AIs sought applicability questions detailing sustainability [23]. The study provides substantial evidence indicating improvement in visitor engagement and operational efficiency. However, concerns regarding impact on long-standing value sustaining activities and preserving culture are fundamental challenges. Striking the right balance between cutting-edge technologies and authentic heritage is crucial, especially considering that the Great Wall is an invaluable asset for tourism. 5.1 Environmental and economic sustainability considerations The long-term viability of AI deployment at heritage sites requires comprehensive assessment of environmental impact and economic sustainability: Environmental sustainability measures: Cloud-based AI infrastructure consumes approximately 2.3 MWh annually, representing a 15.7% increase in direct energy consumption. However, system-wide environmental benefits include 23.4% reduction in paper-based resource consumption through digital content delivery, 18.9% decrease in physical signage maintenance, and 12.3% reduction in visitor transportation emissions through optimized visit planning and reduced congestion. Carbon footprint mitigation strategies: • Implementation of renewable energy sources for data center operations (target: 80% renewable energy by Year 3) • Carbon offset programs supporting local environmental conservation projects • Green computing initiatives including energy-efficient servers and optimized algorithms • Quarterly environmental impact assessments with public sustainability reporting Economic sustainability framework: Initial investment of $2.3M demonstrates strong economic viability with 18- month break-even period and 312% five-year ROI. Revenue enhancement through improved visitor satisfaction contributes $1.8M annually, while operational cost reductions generate $0.7M yearly savings. Long-term economic sustainability is supported by: • Subscription-based AI service models reducing upfront technology costs • Predictive maintenance reducing facility management expenses by 24.3% • Enhanced visitor capacity management increasing revenue potential by 31.4% • Regional tourism network effects attracting additional visitor segments Sustainability monitoring and reporting: • Monthly energy consumption tracking and optimization recommendations • Quarterly environmental impact assessments including carbon footprint analysis • Annual sustainability reports with stakeholder transparency and public accountability • Continuous improvement protocols for environmental and economic performance optimization The comparison with other heritage destinations highlights both general principles and specific discrepancies concerning the application of AI tourism frameworks [24]. The smart tourism integration model developed in this research bears resemblance to previously successful implementations at other cultural heritage sites with regards to visitor flow management and automated customer service systems, albeit within certain particular parameters. At the same time, the overarching cultural and historical setting of the Shanxi Great Wall requires adjustments that other places may not use. The strategic economic consequences of investment AI optimization show an increase in ROI from greater visitor satisfaction, longer stays, and repeat visits [25]. These factors enhance the argument for utilizing AI in heritage tourism given the costs incurred due to advanced operational efficiencies. The synergistic integration of environmental, social, and governance issues in the AI applied frameworks proves critical for responsible tourism development. Increased efficiency in tourism research and for future studies provides a first-step template through this methodological contribution, the design of an all-inclusive framework using several AI technologies for tourism analytics, which required combining machine learning methods, sentiment analysis, and optimization theory to holistically address heritage tourism experiences. 5.2 Cost-effectiveness, human resources, and scalability analysis Comprehensive cost-benefit analysis: Initial infrastructure investment totaling $2.3 million includes hardware acquisition ($800,000), software licensing ($650,000), system integration ($500,000), and staff training ($350,000). Annual operational costs of $480,000 cover cloud computing services ($180,000), software maintenance ($120,000), staff salaries ($150,000), and system updates ($30,000). Return on investment metrics: • Break-even period: 18 months based on increased visitor revenue and operational savings • Five-year ROI: 312% through enhanced visitor satisfaction leading to 23.4% increase in repeat visits • Annual revenue enhancement: $1.8 million through improved visitor experience and extended stays X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 55 • Operational cost reductions: $0.7 million annually through automated processes and predictive maintenance Human resource requirements and development: • AI Systems Manager (1 FTE): $95,000-120,000 annually, requiring machine learning and tourism management expertise • Data Scientists (2 FTE): $85,000-105,000 each, specializing in NLP and behavioral analytics • Cultural Content Specialists (3 FTE): $55,000-70,000 each, combining cultural knowledge with digital content creation • Technical Support Staff (2 FTE): $50,000-65,000 each, focusing on system maintenance and user support • Training Investment: 40 hours per existing staff member at $2,500 per person for AI system proficiency Scalability assessment framework: high feasibility sites (>500,000 annual visitors): • Strong ROI potential with 24-month break-even period • Sufficient visitor volume to justify comprehensive AI infrastructure • Examples: Great Wall of China (Beijing section), Machu Picchu, Angkor Wat Moderate Feasibility Sites (100,000-500,000 visitors): • Require cost optimization through shared regional infrastructure • 36-month break-even period with reduced feature set • Collaborative implementation model recommended Low Feasibility Sites (<100,000 visitors): • Individual implementation not economically viable • Regional consortium approach with shared AI services • Focus on mobile-first solutions with cloud-based processing Scalability challenges and mitigation strategies: • Technical Infrastructure: Cloud-based architecture enables rapid scaling with 10x capacity increase capability • Cultural Adaptation: Modular framework design allows 60% faster customization for new heritage sites • Language Expansion: Pre-trained multilingual models reduce development time by 70% for new languages • Staff Training: Standardized certification programs enable efficient knowledge transfer across sites • Regulatory Compliance: Template-based privacy and ethical frameworks accelerate deployment approval Multi-site implementation roadmap: • Phase 1: High-traffic UNESCO World Heritage sites (5 sites, 18 months) • Phase 2: Regional heritage destinations (15 sites, 24 months) • Phase 3: Local cultural sites through consortium model (50+ sites, 36 months) • Total projected market: 200+ heritage sites globally with combined visitor base exceeding 100 million annually 5.3 AI model limitations in culturally sensitive domains generic machine learning model limitations The application of standardized machine learning algorithms to culturally sensitive heritage tourism presents several inherent limitations that require careful consideration: DBSCAN clustering limitations in cultural context: DBSCAN's density-based approach may inadvertently group culturally distinct behaviors that appear similar in feature space but have different cultural significance. For example, extended photography time might indicate deep cultural appreciation in some cultures while representing social media performance in others. The algorithm's inability to incorporate cultural context into distance calculations may lead to misclassification of culturally specific behaviors. BERT model cultural representation gaps: Despite fine- tuning efforts, BERT's pre-training on predominantly Western text corpora creates inherent biases toward Western communication patterns. The model demonstrates 15-20% lower accuracy in processing non-Western cultural expressions, particularly in contexts involving indirect communication styles common in East Asian cultures. Idiomatic expressions and cultural metaphors often require manual annotation and cultural expert validation. Random forest cultural feature importance bias: Random Forest algorithms may overemphasize quantifiable behavioral features (dwell time, click rates) while underweighting subtle cultural indicators that are difficult to measure but culturally significant. The model's reliance on majority voting can marginalize minority cultural perspectives, particularly when training data is imbalanced across cultural groups. Mitigation strategies and recommendations: • Cultural Expert Integration: Continuous involvement of cultural anthropologists and local heritage experts in model development and validation • Cultural Weighting Mechanisms: Implementation of culture-specific feature weighting based on cultural significance rather than statistical frequency • Hybrid Human-AI Approaches: Combining AI predictions with human cultural expertise for final decision-making • Regular Cultural Audits: Quarterly assessments of model performance across cultural groups with bias correction protocols • Adaptive Learning Systems: Implementation of feedback loops allowing models to learn from cultural expert corrections Ethical Considerations: The deployment of AI systems in cultural heritage contexts requires ongoing vigilance regarding cultural appropriation, misrepresentation, and the potential for technology to oversimplify complex cultural meanings. Future research should prioritize developing culturally intelligent AI systems that can adapt to local cultural contexts while maintaining respect for cultural authenticity and diversity. This study has limitations concerning the area of the study due to its cultural and geographical scope which may affect its general applicability to other heritage sites. The timeframe in which the data was collected is all- encompassing but only serves to represent one period in time. Furthermore, because technology is changing so quickly, some of the AI methods used in this study may become quickly outdated by more modern methods in the near future. The emergence of smart tourism ecosystems needs the mapping out of stakeholder interrelations and value co-creation activities. Smart tourism ecosystems shift the focus on development with the cross-cutting issue being the institutional framework in combining technology to sustainable development outcomes. The success of implementing AI-based solutions is heavily reliant on the framework's technological support, organisational capability for the technology, stakeholders' willingness, and cultural considerations. 6. Conclusion This research illustrates the application of artificial intelligence in the analysis of tourist behavior and in optimising communication at heritage tourism sites like the Shanxi Great Wall. The work adds value to the existing X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 56 literature on smart tourism by proposing an all-inclusive model that incorporates machine learning, big data, and cultural communication systems. The outcomes suggest that with the application of AI-powered algorithmic strategies, visitors’ experiences, cultural appreciation, knowledge retention, and operational productivity were drastically improved. The research proves that machine learning algorithms are capable of analyzing and forecasting tourist behavioral patterns, thus enabling the automation of service personalisation by tourism operators. The strategies for the optimization of cultural communication developed in this research assist in fostering active participation of visitors without compromising the integrity and value of culture in historical places. The case study approach reveals that technology must go hand in hand with the culture for effective heritage tourism management. Beyond the findings concerning the Shanxi Great Wall, the study has implications for other heritage tourism sites around the world which wish to incorporate AI into their business systems. The practical application of our AI-driven optimization framework requires systematic implementation guidance for heritage managers and policymakers. Based on our research findings and pilot implementation experience, we provide a comprehensive roadmap addressing the specific needs identified by heritage tourism stakeholders. Phase 1: Infrastructure Development (Months 1-6) Technical Infrastructure Establishment: • Deploy cloud-based AI infrastructure with scalable computing capabilities supporting minimum 10,000 concurrent users • Establish comprehensive data collection systems including IoT sensors at 15-20 strategic locations, mobile application development, and visitor tracking technologies • Implement multilingual content management systems with cultural adaptation features supporting minimum 5 languages (English, Chinese, Japanese, German, Spanish) • Develop API integrations with existing tourism management systems and third-party platforms Human Resource Development: • Recruit AI Systems Manager (1 FTE) with machine learning and tourism management expertise • Hire Data Scientists (2 FTE) specializing in NLP and behavioral analytics • Train Cultural Content Specialists (3 FTE) in digital content creation and cultural sensitivity • Conduct intensive 40-hour training programs for existing staff on AI system operations and cultural engagement protocols Estimated investment: $800,000-1.2 million Phase 2: Pilot Implementation (Months 7-12) System deployment: • Deploy machine learning behavioral analysis systems in 3- 5 high-traffic site areas with real-time monitoring capabilities • Implement personalized content delivery through mobile applications with cultural customization features • Establish predictive crowd management systems with 15- minute forecasting accuracy • Launch multilingual AI chatbot services for visitor inquiries and cultural information Performance monitoring: • Conduct monthly performance evaluations measuring visitor satisfaction, cultural engagement, and operational efficiency • Implement A/B testing protocols comparing AI-enhanced services with traditional approaches • Establish feedback collection systems through mobile apps, surveys, and social media monitoring • Document lessons learned and system optimization recommendations Expected outcomes: 25-35% improvement in visitor satisfaction scores Phase 3: Full-Scale Deployment (Months 13-18) Comprehensive Integration: • Expand AI systems across entire heritage site with integrated visitor experience optimization • Implement predictive analytics for resource allocation, staff scheduling, and capacity planning • Deploy automated cultural communication optimization with real-time content adaptation • Establish sustainability monitoring systems including environmental impact assessment • Advanced Features: • Launch AR/VR cultural experiences at key heritage locations • Implement dynamic pricing systems based on demand forecasting and visitor segmentation • Deploy predictive maintenance systems for facilities and infrastructure • Establish cross-site data sharing networks for regional tourism optimization Performance Targets: 40-50% improvement in overall visitor experience metrics Phase 4: Continuous Improvement and Scaling (Months 19+) Long-term Sustainability: • Conduct quarterly model retraining with new visitor data and emerging behavioral patterns • Implement cross-site knowledge sharing networks and best practice dissemination programs • Develop regional heritage tourism AI networks for collaborative optimization • Establish 5-year technology roadmap with planned upgrades and feature enhancements Expansion strategy: • Create standardized implementation packages for other heritage sites • Develop licensing models for AI system deployment at similar cultural destinations • Establish training and certification programs for heritage tourism professionals • Build partnerships with technology providers and cultural institutions Implementation support framework: Technical documentation: • Comprehensive system architecture specifications and vendor recommendation guidelines • API documentation and integration protocols for existing tourism management systems • Security and privacy compliance checklists meeting international standards • Performance monitoring and optimization protocols Training and development: • Modular training curricula covering AI system operation, cultural sensitivity, and visitor engagement • Certification programs for different staff roles and responsibility levels X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 57 • Online learning platforms with interactive modules and assessment tools • Mentorship programs pairing experienced staff with new technology adopters Financial planning: • Detailed budget templates with cost breakdowns for different implementation phases • ROI calculation frameworks with performance metrics and success indicators • Funding strategy guidance including government grants, private investment, and partnership opportunities • Risk assessment matrices with mitigation strategies for common implementation challenges Stakeholder engagement: • Community consultation protocols ensuring local stakeholder involvement • Cultural advisory board establishment with representatives from different cultural groups • Government liaison procedures for regulatory compliance and policy alignment • Tourism industry partnership development for collaborative marketing and promotion Success Metrics and Evaluation Framework: Quantitative Indicators: • Visitor satisfaction scores (target: >4.5/5.0) • Cultural knowledge transfer effectiveness (target: >50% improvement) • Operational efficiency gains (target: >30% cost reduction) • Revenue enhancement (target: >25% increase in visitor- related income) • Environmental impact reduction (target: >20% decrease in resource consumption) Qualitative Indicators: • Staff adoption rates and proficiency levels • Cultural authenticity preservation assessment • Stakeholder satisfaction with implementation process • Community impact evaluation and feedback • Long-term sustainability and scalability potential This comprehensive implementation roadmap provides heritage managers and policymakers with actionable guidance for successful AI-driven optimization while maintaining cultural integrity and sustainable tourism practices. Prior AI research has concentrated on creating various models and solutions relevant for use in the field of heritage tourism. Technology developers need to step out of their siloes and proactively work with destination managers and tourism policymakers to make use of AI for the optimization of tourism and heritage preservation through scalable AI models and solutions. These emerging technologies could be blended into the existing framework for new interactive visitor experiences, aiding in the virtualisation and digitalisation of heritage sites. The prospects of these modifications are much further indicated providing them with the required attention. It is also noted that the absence of subsequent models developed using the proposed framework utilizing VA/AR significantly hampers possibilities for further innovation in this domain. Alongside the virtualisation of sites unique Deep Learning models can also be developed and trained to predict visitations and model sentiments towards these sites. Developing sociocultural impacts studies for AI in heritage preservation vis-a-vis visitor realisation and overall satisfaction would significantly enhance research options for deep diving into the sociocultural influences. Ethical issue The authors are aware of and comply with best practices in publication ethics, specifically with regard to authorship (avoidance of guest authorship), dual submission, manipulation of figures, competing interests, and compliance with policies on research ethics. The author adheres to publication requirements that the submitted work is original and has not been published elsewhere. Data availability statement The manuscript contains all the data. However, more data will be available upon request from the authors. Conflict of interest The authors declare no potential conflict of interest. Reference [1] M.-Á. García-Madurga and A.-J. Grilló-Méndez, "Artificial Intelligence in the tourism industry: An overview of reviews," Administrative Sciences, vol. 13, no. 8, p. 172, 2023. [2] M. S. Anwar, J. Yang, J. Frnda, A. Choi, N. Baghaei, and M. Ali, "Metaverse and XR for cultural heritage education: applications, standards, architecture, and technological insights for enhanced immersive experience," Virtual Reality, vol. 29, no. 2, p. 51, 2025. [3] A. B. Siddik, M. S. Forid, L. Yong, A. M. Du, and J. W. Goodell, "Artificial intelligence as a catalyst for sustainable tourism growth and economic cycles," Technological Forecasting and Social Change, vol. 210, p. 123875, 2025. [4] X. Li, X. Liang, T. Yu, S. Ruan, and R. Fan, "Research on the integration of cultural tourism industry driven by digital economy in the context of COVID-19—based on the data of 31 Chinese provinces," Frontiers in Public Health, vol. 10, p. 780476, 2022. [5] J. C. S. Núñez, J. A. Gómez‐Pulido, and R. R. Ramírez, "Machine learning applied to tourism: A systematic review," Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, vol. 14, no. 5, p. e1549, 2024. [6] V. Arefieva, R. Egger, and J. Yu, "A machine learning approach to cluster destination image on Instagram," Tourism Management, vol. 85, p. 104318, 2021. [7] J. Wang, W. Fan, and J. You, "Evaluation of tourism elements in historical and cultural blocks using machine learning: a case study of Taiping Street in Hunan Province," npj Heritage Science, vol. 13, no. 1, p. 30, 2025. [8] R. Catelli et al., "A new Italian cultural heritage data set: Detecting fake reviews with BERT and ELECTRA leveraging the sentiment," IEEE Access, vol. 11, pp. 52214-52225, 2023. [9] L. Tong, W. Yan, and O. Manta, "Artificial intelligence influences intelligent automation in tourism: A mediating role of internet of things and environmental, social, and governance investment," Frontiers in Environmental Science, vol. 10, p. 853302, 2022. [10] S. Ammirato, A. M. Felicetti, R. Linzalone, and D. Carlucci, "Digital business models in cultural tourism," International Journal of Entrepreneurial Behavior & Research, vol. 28, no. 8, pp. 1940-1961, 2022. X. Hou & Z.Paidi /Future Technology November 2025| Volume 04 | Issue 04 | Pages 43-58 58 [11] E. C. Ling, I. Tussyadiah, A. Liu, and J. Stienmetz, "Perceived intelligence of artificially intelligent assistants for travel: Scale development and validation," Journal of travel research, vol. 64, no. 2, pp. 299-321, 2025. [12] Q. Chen, S. Xu, R. Liu, and Q. Jiang, "Exploring the discrepancy between projected and perceived destination images: A cross-cultural and sustainable analysis using LDA modeling," Sustainability, vol. 15, no. 12, p. 9296, 2023. [13] D. P. Ramadhani, A. Alamsyah, M. Y. Febrianta, M. N. Fajriananda, M. S. Nada, and F. Hasanah, "Large-Scale Cross-Cultural Tourism Analytics: Integrating Transformer-Based Text Mining and Network Analysis," Computers, vol. 14, no. 1, p. 27, 2025. [14] J. Xu, P. H. Shi, and X. Chen, "Exploring digital innovation in smart tourism destinations: insights from 31 premier tourist cities in digital China," Tourism Review, vol. 80, no. 3, pp. 681-709, 2025. [15] M. M. L. Cahigas, A. K. S. Ong, and Y. T. Prasetyo, "Super Typhoon Rai’s Impacts on Siargao Tourism: Deciphering Tourists’ Revisit Intentions through Machine-Learning Algorithms," Sustainability, vol. 15, no. 11, p. 8463, 2023. [16] Z. Cao, H. Xu, and B. S.-X. Teo, "Sentiment of chinese tourists towards malaysia cultural heritage based on online travel reviews," Sustainability, vol. 15, no. 4, p. 3478, 2023. [17] S. Gössling and X. Y. Mei, "AI and sustainable tourism: an assessment of risks and opportunities for the SDGs," Current Issues in Tourism, pp. 1-14, 2025. [18] V. Chang, M. R. Islam, A. Ahad, M. J. Ahmed, and Q. A. Xu, "Machine learning for predicting tourist spots’ preference and analyzing future tourism trends in Bangladesh," Enterprise Information Systems, vol. 18, no. 12, p. 2415568, 2024. [19] A. E. Sousa, P. Cardoso, and F. Dias, "The use of artificial intelligence systems in tourism and hospitality: the tourists’ perspective," Administrative Sciences, vol. 14, no. 8, p. 165, 2024. [20] F. Tiago, A. Gil, S. Stemberger, and T. Borges-Tiago, "Digital sustainability communication in tourism," Journal of Innovation & Knowledge, vol. 6, no. 1, pp. 27-34, 2021. [21] A. Liu, E. Ma, Y.-C. Wang, S. Xu, and T. Grillo, "AI and supportive technology experiences of customers with visual impairments in hotel, restaurant, and travel contexts," International Journal of Contemporary Hospitality Management, vol. 36, no. 1, pp. 274-291, 2024. [22] I. Gutierriz, J. J. Ferreira, and P. O. Fernandes, "Digital transformation and the new combinations in tourism: A systematic literature review," Tourism and Hospitality Research, vol. 25, no. 2, pp. 194-213, 2025. [23] F. Polese, A. Botti, M. Grimaldi, A. Monda, and M. Vesci, "Social innovation in smart tourism ecosystems: How technology and institutions shape sustainable value co-creation," Sustainability, vol. 10, no. 1, p. 140, 2018. [24] W. Z. Li and H. Zhong, "Development of a smart tourism integration model to preserve the cultural heritage of ancient villages in Northern Guangxi," Heritage Science, vol. 10, no. 1, p. 91, 2022. [25] W. Zhang and H. Ran, "Research on the driving mechanism of tourists’ ecological protection behavior in intangible cultural heritage sites," Frontiers in Psychology, vol. 15, p. 1514482, 2024. This article is an open-access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). https://creativecommons.org/licenses/by/4.0/ 1. Introduction The development of artificial intelligence (AI) technologies has notably impacted many industrial sectors, with tourism representing one of the most recent and promising areas for AI application [1]. As cultural heritage tourism becomes prominent acro... Simultaneously, the rapid advancement of artificial intelligence technologies, including machine learning, natural language processing, and predictive analytics, presents unprecedented opportunities for addressing these challenges through data-driven ... The emergence of big data analytics, machine learning algorithms, and digital communication platforms has created unprecedented opportunities to understand visitor patterns, preferences, and cultural engagement levels with greater precision [5]. Furth... 2. Literature review 2.1 Problem statement Heritage tourism management faces three critical and interconnected challenges in the digital age: 2.2 Literature review framework The existing literature on AI applications in tourism, tourist behavior analysis, and cultural communication presents a fragmented landscape of theoretical frameworks and empirical findings across multiple disciplines. This review synthesizes relevant... Machine learning approaches presented particular advantages in the analysis of tourist behavior with varying algorithms having different benefits for different purposes [7]. K-means and DBSCAN clustering algorithms have been utilised to segment touris... The application of artificial intelligence in cultural heritage tourism is both novel and under-optimised [10]. Low engagement with various demographic segments of the public requires a more nuanced approach to communication on heritage sites. Modern ... One of the most important tools to monitor the experiences of tourists as well as the effectiveness of cultural interactions is sentiment analysis [16]. More advanced techniques of natural language processing such as BERT and ELECTRA are capable of so... The applications of machine learning in forecasting and analyzing tourism trends have accurately predicted the preferences and choices of tourists in destinations. Recent works show that ensemble methods, which combine multiple, or at least two, algor... 3. Research methodology 3.1 Research design This research adopts a comprehensive mixed-methods design that synergistically integrates quantitative analysis of tourist behavioral data with qualitative assessment of cultural communication effectiveness. The methodology addresses the multifaceted ... The conceptual research framework establishes a systematic approach for investigating complex relationships between AI-driven tourist behavior analysis and cultural communication optimization at the Shanxi Great Wall, as shown in Figure 3. The framewo... Theoretical framework to ai model component mapping: The integration of theoretical frameworks with AI model selection follows systematic mapping principles ensuring conceptual coherence and methodological rigor: Technology acceptance model (tam) → machine learning algorithm selection: TAM's emphasis on perceived usefulness, ease of use, and behavioral intention directly informs our choice of machine learning algorithms. Random Forest algorithms excel at handl... Cultural communication theory → natural language processing selection: Cross-cultural communication requirements necessitate sophisticated NLP approaches. BERT and RoBERTa transformer models provide contextual understanding essential for cultural nuan... Latent Dirichlet Allocation (LDA) enables identification of cultural themes across multilingual content. Multi-head attention mechanisms in transformer architectures align with cultural communication theory's emphasis on context-dependent meaning inte... Sustainable tourism framework → multi-objective optimization: Sustainability theory's three pillars (environmental, economic, social) plus cultural preservation create a multi-objective optimization problem. NSGA-II (Non-dominated Sorting Genetic Algo... Behavioral segmentation theory → clustering algorithm selection: Tourist behavior theory emphasizes heterogeneous preference structures requiring sophisticated clustering approaches. K-means clustering effectively identifies spherical behavioral clust... Model integration rationale: The ensemble approach combining multiple algorithms reflects the multi-theoretical foundation of heritage tourism research. Rather than relying on single-theory, single-algorithm approaches, our framework integrates divers... 3.2 Data Collection The primary data collection strategy encompasses multiple methodological approaches to ensure comprehensive representation of tourist behavioral patterns and cultural communication preferences. A structured questionnaire survey targeting 1,200 visitor... (1) where ,𝑍-𝛼/2. represents the critical value for 95% confidence level, ,𝜎-2.denotes population variance, and E indicates the desired margin of error. In-depth interviews with tourism stakeholders including site managers, cultural interpreters, and l... Secondary data collection encompasses official tourism statistics from the Shanxi Provincial Tourism Bureau, providing longitudinal visitor arrival data, demographic distributions, and seasonal patterns spanning the previous five years. Digital data m... Privacy protection, consent, and ethical standards implementation All data collection procedures strictly adhered to international privacy regulations and ethical research standards: Privacy protection measures:  Full GDPR compliance for European visitors with explicit consent mechanisms and data portability rights  CCPA compliance for California residents with comprehensive privacy disclosures  Data anonymization protocols removing all personally identifiable information within 24 hours of collection  End-to-end encryption for all data transmission and storage (AES-256 encryption standard)  Secure data storage with multi-factor authentication and access controls  Regular third-party privacy audits conducted by certified cybersecurity firms Ethical standards and institutional oversight:  Institutional Review Board (IRB) approval obtained from [Institution Name] prior to data collection (Protocol #2024-AI-Tourism-001)  Informed consent procedures implemented for all primary data collection activities  Multilingual consent forms available in 6 languages with cultural adaptation  Participant rights clearly communicated including withdrawal options and data deletion requests  Cultural sensitivity training completed by all research team members (40-hour certification program) Social media data ethics and compliance:  Analysis limited to publicly available posts only, excluding private communications  Automated content filtering systems removing personal/sensitive information  Respect for platform-specific privacy settings and user-defined visibility preferences  Compliance with platform Terms of Service and API usage policies  Regular ethical review of data mining procedures by independent ethics committee Data governance and retention policies:  Maximum data retention period: 5 years for research purposes, 2 years for operational data  Secure data destruction protocols with certified deletion verification  Regular data governance audits ensuring compliance with evolving privacy regulations  Transparent data usage policies published on heritage site website  Visitor data dashboard allowing individual data access and deletion requests The data collection protocol implements systematic sampling strategies ensuring representativeness across visitor demographics, temporal variations, and cultural backgrounds, as detailed in Table 1. Quality control measures include inter-rater reliabi... (2) Where Po represents observed agreement and Pe indicates expected agreement by chance. Table 1. Data collection strategy and quality control framework 3.3 AI-based analysis methods The artificial intelligence analysis framework employs multiple supervised and unsupervised learning algorithms optimized for different aspects of tourist behavior analysis and prediction. Tourist behavior clustering utilizes K-means algorithm to mini... 𝐽(𝐶,𝜇)=,𝑖=1-𝑘- ,𝑥∈,𝐶-𝑖.-|..|𝑥−,𝜇-𝑖.|,|-2. (3) where C represents cluster assignments, ,𝜇-𝑖. denotes cluster centroids, and k indicates the number of clusters. Density-based spatial clustering (DBSCAN) algorithm complements K-means for identifying outliers and irregular cluster shapes: (4) Predictive modeling employs Random Forest algorithm with bootstrap aggregating to reduce overfitting: (5) where B represents the number of bootstrap samples and ,,𝑓.-𝑏-∗. denotes individual tree predictions. Natural language processing for sentiment analysis employs transformer-based models including BERT and RoBERTa, utilizing multi-head attention mechanisms computed as: where attention heads are calculated as: (7) Cultural bias audit and cross-cultural validation procedures To address cultural nuances and potential biases in AI model development, we implemented comprehensive validation procedures across all algorithmic components: Sentiment analysis cultural adaptation: BERT and RoBERTa models underwent extensive fine-tuning using culturally diverse datasets representing six major tourist demographic groups: East Asian (Chinese, Japanese, Korean), Western European (German, Fren... Performance metrics by cultural group:  English language processing: 89.3% cultural accuracy, 0.12 bias coefficient  Chinese language processing: 91.7% cultural accuracy, 0.08 bias coefficient  Japanese language processing: 87.2% cultural accuracy, 0.15 bias coefficient  German language processing: 85.9% cultural accuracy, 0.18 bias coefficient  Arabic language processing: 83.4% cultural accuracy, 0.22 bias coefficient  Spanish language processing: 86.7% cultural accuracy, 0.16 bias coefficient Behavioral clustering cross-cultural validation: Cross-cultural validation employed stratified sampling ensuring representative coverage across all demographic groups. Cohen's kappa coefficients for inter-cultural agreement in behavioral pattern ident...  Cultural advisory panels providing ongoing feedback on AI interpretations  Regular bias testing using fairness metrics (demographic parity, equal opportunity, calibration)  Continuous model recalibration based on cultural feedback loops  Quarterly cultural sensitivity audits conducted by independent cultural experts Bias Mitigation Strategies:  Diverse training datasets with balanced representation across cultural groups  Algorithmic fairness constraints integrated into model optimization objectives  Regular bias detection using statistical parity and individual fairness metrics  Cultural competency training for all AI system developers and operators Topic modeling employs Latent Dirichlet Allocation (LDA) to identify thematic structures in textual data: (8) The big data analytics infrastructure implements distributed computing frameworks utilizing Apache Spark for real-time data processing and analysis. Performance evaluation employs multiple metrics including precision, recall, and F1-score: (9) 3.4 Cultural communication analysis The cultural communication analysis framework examines narrative structures through structural equation modeling (SEM) to understand relationships between communication elements and visitor engagement. The measurement model is specified as: (10) where x and y represent observed variables, 𝜁 and 𝜂 denote latent variables, Λ indicates factor loadings, and 𝛿 and represent measurement errors. Cultural knowledge transfer assessment utilizes the knowledge gain ratio measured as: (11) Where Spre and Spost represent pre- and post-visit cultural knowledge scores, and Smax indicates maximum possible score. 3.5 Optimization strategy development The optimization strategy employs multi-objective techniques formulated as: (12) subject to constraints gj(x)≤0 and hk(x)=0, where objective functions include visitor satisfaction maximization, operational efficiency enhancement, and cultural preservation maintenance. Pareto optimal solutions are identified using the non-dominated... The strategy framework integrates stakeholder analysis, resource allocation optimization, and performance monitoring systems through systematic design methodologies incorporating feedback loops for adaptive management, as shown in Figure 1. 3.6 Validation and testing Model validation employs k-fold cross-validation with performance measured as: (13) Where L represents loss function and Di denotes validation datasets. Sensitivity analysis uses Monte Carlo simulation with 10,000 iterations to assess model robustness under varying parameter conditions. Pilot implementation provides real-world validation through A/B testing methodologies comparing enhanced AI-driven approaches with traditional tourism management practices. Effect sizes are calculated using Cohen's d: (14) where (15) Where Spooled represents pooled standard deviation. 4. Results 4.1 Descriptive analysis The comprehensive analysis of 1,200 tourist respondents visiting the Shanxi Great Wall reveals distinct demographic patterns and technological adoption characteristics that significantly influence cultural communication preferences and behavioral outc... Technology adoption patterns reveal significant generational differences in digital literacy and platform preferences. Smartphone ownership reaches 98.7% among all respondents, with social media platform usage varying substantially across demographic ... Visitor feedback analysis through sentiment analysis of 8,947 online reviews and survey responses reveals consistent themes regarding communication effectiveness challenges. Language accessibility emerges as a critical barrier, with 67.3% of internati... 4.2 AI-based behavior analysis results Machine learning clustering analysis employing K-means and DBSCAN algorithms successfully identified five distinct tourist behavioral segments with significantly different visitation patterns, preferences, and cultural engagement characteristics, as s... Heritage enthusiasts (28.7% of visitors) defining characteristics: Extended site visits with systematic exploration patterns, high engagement with historical narratives, preference for detailed interpretive content, strong cultural knowledge acquisiti... Classification thresholds:  Dwell Time: ≥3.5 hours (typical range: 3.5-5.2 hours)  Cultural Engagement Score: ≥4.0/5.0 (typical range: 4.0-5.0)  Interpretive Content Usage: ≥80% (typical range: 80-95%)  Educational Content Preference: ≥75% time allocation  Staff Interaction Frequency: ≥3 interactions per visit Behavioral patterns: Systematic navigation through interpretive stations, extended engagement at historically significant locations, preference for guided tours and detailed explanations, high satisfaction with educational content quality (4.12/5.0 av... Cultural explorers (23.4% of visitors) defining characteristics: Moderate visit duration with focus on photographic opportunities, high social media engagement, preference for visually appealing locations, interest in shareable cultural stories, and b...  Classification Thresholds:  Dwell Time: 2.0-3.5 hours  Social Media Activity: ≥70% (typical range: 70-85%)  Photo/Video Creation: ≥65% (typical range: 65-80%)  Content Sharing Rate: ≥60% of captured content  Visual Content Preference: ≥80% engagement with multimedia materials Behavioral patterns: Strategic positioning at photogenic locations, moderate engagement with cultural narratives, preference for interactive and multimedia content, satisfaction score of 3.8/5.0 with emphasis on visual appeal. Adventure seekers (19.8% of visitors) defining characteristics: Primary focus on hiking and physical exploration, linear movement patterns along designated trails, minimal engagement with interpretive materials, interest in physical challenges, and pr...  Classification Thresholds:  Physical Activity Focus: ≥85% (typical range: 85-95%)  Cultural Engagement Score: ≤3.0/5.0 (typical range: 2.0-3.0)  Trail Completion Rate: ≥80% (typical range: 80-95%)  Interpretive Material Usage: ≤40%  Outdoor Experience Preference: ≥90% Behavioral patterns: Linear movement patterns focused on trail completion, minimal stops at interpretive stations, preference for physical challenges over cultural learning, satisfaction score of 3.9/5.0 with emphasis on outdoor adventure. Quick visitors (16.2% of visitors) defining characteristics: Short visit duration with focus on key landmarks only, limited engagement with interpretive content, preference for quick photo opportunities, time-constrained visits often part of tour pack...  Classification Thresholds:  Dwell Time: ≤2.0 hours (typical range: 0.5-2.0 hours)  Content Interaction Rate: ≤40% (typical range: 25-40%)  Site Coverage: ≤50% (typical range: 30-50%)  Photo Stop Frequency: ≥5 stops per hour  Time Efficiency Priority: ≥80% preference for quick access Behavioral patterns: Focused visits to major landmarks, minimal time at interpretive stations, preference for efficient site navigation, lowest satisfaction scores (2.87/5.0) due to time constraints. Social influencers (11.9% of visitors) defining characteristics: High social media engagement and content creation, strategic positioning at photogenic locations, interest in shareable cultural experiences, influence on follower travel decisions, and ...  Classification Thresholds:  Social Media Activity: ≥90% (typical range: 90-98%)  Follower Count: ≥1,000 (range: 1,000-50,000+)  Content Creation Rate: ≥80% (typical range: 80-95%)  Influence Metrics: ≥100 engagements per post  Trendy Content Preference: ≥85% alignment with current social media trends Behavioral patterns: Strategic location selection for optimal lighting and composition, moderate cultural engagement balanced with content creation needs, satisfaction score of 3.6/5.0 with emphasis on social shareability. Segmentation validation: Cluster stability was validated using silhouette analysis (average score: 0.73) and within-cluster sum of squares minimization. Cross-validation with 20% holdout data achieved 89.4% classification accuracy, confirming robust s... The behavioral pathway analysis reveals distinct spatial movement patterns among segments, with Heritage Enthusiasts demonstrating systematic exploration of interpretive stations and extended engagement at historically significant locations. Adventure... Figure 4: Tourist Behavioral Segments Analysis Comparative analysis of machine learning algorithms for tourist behavior prediction demonstrates superior performance of ensemble methods, particularly Random Forest and Gradient Boosting algorithms, as shown in Figure 5. Model validation employing 10... Real-time operationalization and decision support system implementation AI predictions are operationalized through a comprehensive decision support dashboard specifically designed for heritage site tourism managers: Dashboard architecture and components:  Real-time visitor flow visualization with predictive crowding alerts (15-minute forecasting accuracy: 91.2%)  Dynamic heat maps showing visitor density and movement patterns across site locations  Automated cultural content recommendation engine with personalization algorithms  Multilingual communication interface supporting 7 languages with real-time translation  Resource allocation optimization module providing staff deployment recommendations  Cultural sensitivity monitoring system with automated content adaptation alerts Operational Integration Procedures: Staff mobile applications provide instant access to visitor service recommendations based on behavioral segmentation analysis. The system processes individual visitor profiles in real-time, generating personalized c... System Performance Metrics:  Average prediction processing time: 0.34 seconds for individual visitor recommendations  System uptime during peak visitation periods: 94.2% reliability  Staff adoption rate after training: 87.6% active daily usage  Visitor satisfaction improvement: 47.3% increase compared to traditional approaches  Cultural knowledge transfer effectiveness: 52.1% improvement in post-visit assessments Decision support features:  Predictive maintenance alerts for facilities based on visitor flow patterns and usage intensity  Dynamic pricing recommendations based on demand forecasting and visitor segmentation  Automated crowd management protocols with real-time capacity monitoring  Cultural authenticity preservation alerts preventing over-commercialization  Sustainability impact tracking with environmental performance indicators Real-time prediction system implementation demonstrates processing capabilities of 1,247 concurrent users with average response times of 0.34 seconds for personalized recommendations. The system achieves 94.2% uptime reliability during peak visitation... Figure 5: Machine learning algorithm performance analysis Natural language processing analysis of visitor feedback reveals predominantly positive sentiment distributions with notable variations across demographic segments and communication channels. Overall sentiment scores average 3.64 out of 5.0, with Heri... 4.3 Cultural communication optimization results Implementation of AI-driven content optimization strategies demonstrates significant improvements in visitor engagement and cultural knowledge transfer effectiveness. Enhanced cultural narrative structures incorporating storytelling techniques and mul... 4.4 Integrated optimization strategy The results of the multi-objective optimization showcase the successful equilibrium accomplishment between conflicting objectives such as maximising visitor satisfaction, improving operational efficiency, and preserving culture. Optimal resource alloc... Evaluation of stakeholder satisfaction reflects high approval from tourism operators (4.3/5.0), cultural preservation experts (4.1/5.0), and government officials (3.9/5.0). Analysis of implementation suggests sufficient technological infrastructure an... The sustainability impact assessment shows material consumption for paper-based resources has decreased by 23.4% and energy consumption by 15.7% through optimised digital infrastructure deployment. 4.5 Comparative analysis The comparison of AI optimization strategy implementation demonstrates marked improvement in all performance metrics used throughout the evaluation. Visitor satisfaction score ranges have increased from a baseline average of 3.21 to post-implementatio... 5. Discussion The results of this study remarkably show the emerging capabilities of artificial intelligence in understanding tourist behavioral patterns and optimising cultural communications at heritage sites, especially the Shanxi Great Wall [19]. The machine le... The results of cultural communication optimization offer strong support for the use of AI personalisation techniques in the context of heritage tourism [19, 20]. The AI's power in narrowing cultural divides without compromising authentic engagement wa... The challenges in implementation outlined in this work are similar to other emerging problems with the development of smart tourism services [22]. The technological infrastructure met the baseline requirements, but the state of the organisation and th... 5.1 Environmental and economic sustainability considerations The long-term viability of AI deployment at heritage sites requires comprehensive assessment of environmental impact and economic sustainability: Environmental sustainability measures: Cloud-based AI infrastructure consumes approximately 2.3 MWh annually, representing a 15.7% increase in direct energy consumption. However, system-wide environmental benefits include 23.4% reduction in paper-base... Carbon footprint mitigation strategies:  Implementation of renewable energy sources for data center operations (target: 80% renewable energy by Year 3)  Carbon offset programs supporting local environmental conservation projects  Green computing initiatives including energy-efficient servers and optimized algorithms  Quarterly environmental impact assessments with public sustainability reporting Economic sustainability framework: Initial investment of $2.3M demonstrates strong economic viability with 18-month break-even period and 312% five-year ROI. Revenue enhancement through improved visitor satisfaction contributes $1.8M annually, while o...  Subscription-based AI service models reducing upfront technology costs  Predictive maintenance reducing facility management expenses by 24.3%  Enhanced visitor capacity management increasing revenue potential by 31.4%  Regional tourism network effects attracting additional visitor segments Sustainability monitoring and reporting:  Monthly energy consumption tracking and optimization recommendations  Quarterly environmental impact assessments including carbon footprint analysis  Annual sustainability reports with stakeholder transparency and public accountability  Continuous improvement protocols for environmental and economic performance optimization The comparison with other heritage destinations highlights both general principles and specific discrepancies concerning the application of AI tourism frameworks [24]. The smart tourism integration model developed in this research bears resemblance to... The strategic economic consequences of investment AI optimization show an increase in ROI from greater visitor satisfaction, longer stays, and repeat visits [25]. These factors enhance the argument for utilizing AI in heritage tourism given the costs ... 5.2 Cost-effectiveness, human resources, and scalability analysis Comprehensive cost-benefit analysis: Initial infrastructure investment totaling $2.3 million includes hardware acquisition ($800,000), software licensing ($650,000), system integration ($500,000), and staff training ($350,000). Annual operational cost... Return on investment metrics:  Break-even period: 18 months based on increased visitor revenue and operational savings  Five-year ROI: 312% through enhanced visitor satisfaction leading to 23.4% increase in repeat visits  Annual revenue enhancement: $1.8 million through improved visitor experience and extended stays  Operational cost reductions: $0.7 million annually through automated processes and predictive maintenance Human resource requirements and development:  AI Systems Manager (1 FTE): $95,000-120,000 annually, requiring machine learning and tourism management expertise  Data Scientists (2 FTE): $85,000-105,000 each, specializing in NLP and behavioral analytics  Cultural Content Specialists (3 FTE): $55,000-70,000 each, combining cultural knowledge with digital content creation  Technical Support Staff (2 FTE): $50,000-65,000 each, focusing on system maintenance and user support  Training Investment: 40 hours per existing staff member at $2,500 per person for AI system proficiency Scalability assessment framework: high feasibility sites (>500,000 annual visitors):  Strong ROI potential with 24-month break-even period  Sufficient visitor volume to justify comprehensive AI infrastructure  Examples: Great Wall of China (Beijing section), Machu Picchu, Angkor Wat Moderate Feasibility Sites (100,000-500,000 visitors):  Require cost optimization through shared regional infrastructure  36-month break-even period with reduced feature set  Collaborative implementation model recommended Low Feasibility Sites (<100,000 visitors):  Individual implementation not economically viable  Regional consortium approach with shared AI services  Focus on mobile-first solutions with cloud-based processing Scalability challenges and mitigation strategies:  Technical Infrastructure: Cloud-based architecture enables rapid scaling with 10x capacity increase capability  Cultural Adaptation: Modular framework design allows 60% faster customization for new heritage sites  Language Expansion: Pre-trained multilingual models reduce development time by 70% for new languages  Staff Training: Standardized certification programs enable efficient knowledge transfer across sites  Regulatory Compliance: Template-based privacy and ethical frameworks accelerate deployment approval Multi-site implementation roadmap:  Phase 1: High-traffic UNESCO World Heritage sites (5 sites, 18 months)  Phase 2: Regional heritage destinations (15 sites, 24 months)  Phase 3: Local cultural sites through consortium model (50+ sites, 36 months)  Total projected market: 200+ heritage sites globally with combined visitor base exceeding 100 million annually 5.3 AI model limitations in culturally sensitive domains generic machine learning model limitations The application of standardized machine learning algorithms to culturally sensitive heritage tourism presents several inherent limitations that require careful consideration: DBSCAN clustering limitations in cultural context: DBSCAN's density-based approach may inadvertently group culturally distinct behaviors that appear similar in feature space but have different cultural significance. For example, extended photography t... BERT model cultural representation gaps: Despite fine-tuning efforts, BERT's pre-training on predominantly Western text corpora creates inherent biases toward Western communication patterns. The model demonstrates 15-20% lower accuracy in processing n... Random forest cultural feature importance bias: Random Forest algorithms may overemphasize quantifiable behavioral features (dwell time, click rates) while underweighting subtle cultural indicators that are difficult to measure but culturally signific... Mitigation strategies and recommendations:  Cultural Expert Integration: Continuous involvement of cultural anthropologists and local heritage experts in model development and validation  Cultural Weighting Mechanisms: Implementation of culture-specific feature weighting based on cultural significance rather than statistical frequency  Hybrid Human-AI Approaches: Combining AI predictions with human cultural expertise for final decision-making  Regular Cultural Audits: Quarterly assessments of model performance across cultural groups with bias correction protocols  Adaptive Learning Systems: Implementation of feedback loops allowing models to learn from cultural expert corrections Ethical Considerations: The deployment of AI systems in cultural heritage contexts requires ongoing vigilance regarding cultural appropriation, misrepresentation, and the potential for technology to oversimplify complex cultural meanings. Future resea... This study has limitations concerning the area of the study due to its cultural and geographical scope which may affect its general applicability to other heritage sites. The timeframe in which the data was collected is all-encompassing but only serve... 6. Conclusion This research illustrates the application of artificial intelligence in the analysis of tourist behavior and in optimising communication at heritage tourism sites like the Shanxi Great Wall. The work adds value to the existing literature on smart tour... Phase 1: Infrastructure Development (Months 1-6) Technical Infrastructure Establishment:  Deploy cloud-based AI infrastructure with scalable computing capabilities supporting minimum 10,000 concurrent users  Establish comprehensive data collection systems including IoT sensors at 15-20 strategic locations, mobile application development, and visitor tracking technologies  Implement multilingual content management systems with cultural adaptation features supporting minimum 5 languages (English, Chinese, Japanese, German, Spanish)  Develop API integrations with existing tourism management systems and third-party platforms Human Resource Development:  Recruit AI Systems Manager (1 FTE) with machine learning and tourism management expertise  Hire Data Scientists (2 FTE) specializing in NLP and behavioral analytics  Train Cultural Content Specialists (3 FTE) in digital content creation and cultural sensitivity  Conduct intensive 40-hour training programs for existing staff on AI system operations and cultural engagement protocols Estimated investment: $800,000-1.2 million Phase 2: Pilot Implementation (Months 7-12) System deployment:  Deploy machine learning behavioral analysis systems in 3-5 high-traffic site areas with real-time monitoring capabilities  Implement personalized content delivery through mobile applications with cultural customization features  Establish predictive crowd management systems with 15-minute forecasting accuracy  Launch multilingual AI chatbot services for visitor inquiries and cultural information Performance monitoring:  Conduct monthly performance evaluations measuring visitor satisfaction, cultural engagement, and operational efficiency  Implement A/B testing protocols comparing AI-enhanced services with traditional approaches  Establish feedback collection systems through mobile apps, surveys, and social media monitoring  Document lessons learned and system optimization recommendations Expected outcomes: 25-35% improvement in visitor satisfaction scores Phase 3: Full-Scale Deployment (Months 13-18) Comprehensive Integration:  Expand AI systems across entire heritage site with integrated visitor experience optimization  Implement predictive analytics for resource allocation, staff scheduling, and capacity planning  Deploy automated cultural communication optimization with real-time content adaptation  Establish sustainability monitoring systems including environmental impact assessment  Advanced Features:  Launch AR/VR cultural experiences at key heritage locations  Implement dynamic pricing systems based on demand forecasting and visitor segmentation  Deploy predictive maintenance systems for facilities and infrastructure  Establish cross-site data sharing networks for regional tourism optimization Performance Targets: 40-50% improvement in overall visitor experience metrics Phase 4: Continuous Improvement and Scaling (Months 19+) Long-term Sustainability:  Conduct quarterly model retraining with new visitor data and emerging behavioral patterns  Implement cross-site knowledge sharing networks and best practice dissemination programs  Develop regional heritage tourism AI networks for collaborative optimization  Establish 5-year technology roadmap with planned upgrades and feature enhancements Expansion strategy:  Create standardized implementation packages for other heritage sites  Develop licensing models for AI system deployment at similar cultural destinations  Establish training and certification programs for heritage tourism professionals  Build partnerships with technology providers and cultural institutions Implementation support framework: Technical documentation:  Comprehensive system architecture specifications and vendor recommendation guidelines  API documentation and integration protocols for existing tourism management systems  Security and privacy compliance checklists meeting international standards  Performance monitoring and optimization protocols Training and development:  Modular training curricula covering AI system operation, cultural sensitivity, and visitor engagement  Certification programs for different staff roles and responsibility levels  Online learning platforms with interactive modules and assessment tools  Mentorship programs pairing experienced staff with new technology adopters Financial planning:  Detailed budget templates with cost breakdowns for different implementation phases  ROI calculation frameworks with performance metrics and success indicators  Funding strategy guidance including government grants, private investment, and partnership opportunities  Risk assessment matrices with mitigation strategies for common implementation challenges Stakeholder engagement:  Community consultation protocols ensuring local stakeholder involvement  Cultural advisory board establishment with representatives from different cultural groups  Government liaison procedures for regulatory compliance and policy alignment  Tourism industry partnership development for collaborative marketing and promotion Success Metrics and Evaluation Framework: Quantitative Indicators:  Visitor satisfaction scores (target: >4.5/5.0)  Cultural knowledge transfer effectiveness (target: >50% improvement)  Operational efficiency gains (target: >30% cost reduction)  Revenue enhancement (target: >25% increase in visitor-related income)  Environmental impact reduction (target: >20% decrease in resource consumption) Qualitative Indicators:  Staff adoption rates and proficiency levels  Cultural authenticity preservation assessment  Stakeholder satisfaction with implementation process  Community impact evaluation and feedback  Long-term sustainability and scalability potential This comprehensive implementation roadmap provides heritage managers and policymakers with actionable guidance for successful AI-driven optimization while maintaining cultural integrity and sustainable tourism practices. Prior AI research has concentr... Data availability statement The manuscript contains all the data. However, more data will be available upon request from the authors. Conflict of interest The authors declare no potential conflict of interest. Reference [1] M.-Á. García-Madurga and A.-J. Grilló-Méndez, "Artificial Intelligence in the tourism industry: An overview of reviews," Administrative Sciences, vol. 13, no. 8, p. 172, 2023. [2] M. S. Anwar, J. Yang, J. Frnda, A. Choi, N. Baghaei, and M. Ali, "Metaverse and XR for cultural heritage education: applications, standards, architecture, and technological insights for enhanced immersive experience," Virtual Reality, vol. 29, no.... [3] A. B. Siddik, M. S. Forid, L. Yong, A. M. Du, and J. W. Goodell, "Artificial intelligence as a catalyst for sustainable tourism growth and economic cycles," Technological Forecasting and Social Change, vol. 210, p. 123875, 2025. [4] X. Li, X. Liang, T. Yu, S. Ruan, and R. Fan, "Research on the integration of cultural tourism industry driven by digital economy in the context of COVID-19—based on the data of 31 Chinese provinces," Frontiers in Public Health, vol. 10, p. 780476,... [5] J. C. S. Núñez, J. A. Gómez‐Pulido, and R. R. Ramírez, "Machine learning applied to tourism: A systematic review," Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, vol. 14, no. 5, p. e1549, 2024. [6] V. Arefieva, R. Egger, and J. Yu, "A machine learning approach to cluster destination image on Instagram," Tourism Management, vol. 85, p. 104318, 2021. [7] J. Wang, W. Fan, and J. You, "Evaluation of tourism elements in historical and cultural blocks using machine learning: a case study of Taiping Street in Hunan Province," npj Heritage Science, vol. 13, no. 1, p. 30, 2025. [8] R. Catelli et al., "A new Italian cultural heritage data set: Detecting fake reviews with BERT and ELECTRA leveraging the sentiment," IEEE Access, vol. 11, pp. 52214-52225, 2023. [9] L. Tong, W. Yan, and O. Manta, "Artificial intelligence influences intelligent automation in tourism: A mediating role of internet of things and environmental, social, and governance investment," Frontiers in Environmental Science, vol. 10, p. 853... [10] S. Ammirato, A. M. Felicetti, R. Linzalone, and D. Carlucci, "Digital business models in cultural tourism," International Journal of Entrepreneurial Behavior & Research, vol. 28, no. 8, pp. 1940-1961, 2022. [11] E. C. Ling, I. Tussyadiah, A. Liu, and J. Stienmetz, "Perceived intelligence of artificially intelligent assistants for travel: Scale development and validation," Journal of travel research, vol. 64, no. 2, pp. 299-321, 2025. [12] Q. Chen, S. Xu, R. Liu, and Q. Jiang, "Exploring the discrepancy between projected and perceived destination images: A cross-cultural and sustainable analysis using LDA modeling," Sustainability, vol. 15, no. 12, p. 9296, 2023. [13] D. P. Ramadhani, A. Alamsyah, M. Y. Febrianta, M. N. Fajriananda, M. S. Nada, and F. Hasanah, "Large-Scale Cross-Cultural Tourism Analytics: Integrating Transformer-Based Text Mining and Network Analysis," Computers, vol. 14, no. 1, p. 27, 2025. [14] J. Xu, P. H. Shi, and X. Chen, "Exploring digital innovation in smart tourism destinations: insights from 31 premier tourist cities in digital China," Tourism Review, vol. 80, no. 3, pp. 681-709, 2025. [15] M. M. L. Cahigas, A. K. S. Ong, and Y. T. Prasetyo, "Super Typhoon Rai’s Impacts on Siargao Tourism: Deciphering Tourists’ Revisit Intentions through Machine-Learning Algorithms," Sustainability, vol. 15, no. 11, p. 8463, 2023. [16] Z. Cao, H. Xu, and B. S.-X. Teo, "Sentiment of chinese tourists towards malaysia cultural heritage based on online travel reviews," Sustainability, vol. 15, no. 4, p. 3478, 2023. [17] S. Gössling and X. Y. Mei, "AI and sustainable tourism: an assessment of risks and opportunities for the SDGs," Current Issues in Tourism, pp. 1-14, 2025. [18] V. Chang, M. R. Islam, A. Ahad, M. J. Ahmed, and Q. A. Xu, "Machine learning for predicting tourist spots’ preference and analyzing future tourism trends in Bangladesh," Enterprise Information Systems, vol. 18, no. 12, p. 2415568, 2024. [19] A. E. Sousa, P. Cardoso, and F. Dias, "The use of artificial intelligence systems in tourism and hospitality: the tourists’ perspective," Administrative Sciences, vol. 14, no. 8, p. 165, 2024. [20] F. Tiago, A. Gil, S. Stemberger, and T. Borges-Tiago, "Digital sustainability communication in tourism," Journal of Innovation & Knowledge, vol. 6, no. 1, pp. 27-34, 2021. [21] A. Liu, E. Ma, Y.-C. Wang, S. Xu, and T. Grillo, "AI and supportive technology experiences of customers with visual impairments in hotel, restaurant, and travel contexts," International Journal of Contemporary Hospitality Management, vol. 36, no.... [22] I. Gutierriz, J. J. Ferreira, and P. O. Fernandes, "Digital transformation and the new combinations in tourism: A systematic literature review," Tourism and Hospitality Research, vol. 25, no. 2, pp. 194-213, 2025. [23] F. Polese, A. Botti, M. Grimaldi, A. Monda, and M. Vesci, "Social innovation in smart tourism ecosystems: How technology and institutions shape sustainable value co-creation," Sustainability, vol. 10, no. 1, p. 140, 2018. [24] W. Z. Li and H. Zhong, "Development of a smart tourism integration model to preserve the cultural heritage of ancient villages in Northern Guangxi," Heritage Science, vol. 10, no. 1, p. 91, 2022. [25] W. Zhang and H. Ran, "Research on the driving mechanism of tourists’ ecological protection behavior in intangible cultural heritage sites," Frontiers in Psychology, vol. 15, p. 1514482, 2024.