Introduction

Architectural heritage serves as a crucial vessel for cultural identity, embodying tangible historical assets and vital cultural resources essential for artistic development1. The ICOMOS Charter (1978) provides a foundational definition, characterizing architectural heritage as a material asset possessing historical, artistic, or social value derived from its architectural style, typology, and environmental context2,3.

In recent decades, this invaluable heritage has faced growing threats from natural disasters4, overtourism5, and rapid urbanization6, underscoring the critical need for advanced methodologies in documentation and preservation7. Consequently, virtual technologies (VR, AR, MR) have become prominent tools within architectural heritage research. They offer multisensory 3D experiences and sophisticated digital management approaches, and are now widely used by scholars8,9. Their applications extend beyond digital preservation10, restoration11, and management12 to encompass education13, tourism14, and visitor experiences15.

While the adoption of virtual technologies in architectural heritage conservation has surged, it has also generated a vast, diverse, and increasingly fragmented body of literature. This rapid growth poses a challenge. Despite many studies, there is still no comprehensive synthesis that tracks developments, identifies current and emerging challenges, and maps future research needs in this field. Existing reviews often focus on specific technologies16, narrower application scopes17, and many do not cover the most recent rapid developments18.

This paper addresses this critical gap through a systematic review designed to provide new insights in three key dimensions: (1) synthesizing evolving trends: Identifying and analyzing the latest trends, including technological convergence (e.g., VR/AR/MR integration) and emerging paradigms. (2) Diagnosing future challenges: Systematically identifying pressing challenges beyond technical limitations, encompassing socio-cultural, ethical, and sustainability concerns. And, (3) mapping critical research needs: Explicitly defining priority research needs to provide a roadmap for future work. Our contribution lies in delivering an integrated, critical assessment that clarifies the current state, identifies key challenges, and outlines research directions for the future.

Section 2 details our systematic review methodology. Section 3 synthesizes recent advances in virtual technology applications for architectural heritage, structured by our quantitative insights. Section 4 integrates this analysis, discussing the current state, critical challenges, and emerging research opportunities.

Methods

A systematic review methodology is employed to gain comprehensive insights into research trends within the field and to identify future potential and challenges of virtual technologies in architectural heritage conservation (Fig. 1). This approach, previously utilized by Yin, et al. and Zhong, et al.19,20, has demonstrated its efficacy in mitigating biased conclusions and subjective interpretations. Moreover, Greene, et al. delineated five essential characteristics of systematic reviews21: (1) triangulation: enabling cross-corroboration between quantitative and qualitative data; (2) complementarity: utilizing results from one method to elucidate or enhance the credibility of another; (3) development: sequencing research methods to optimize the utilization of preceding findings; (4) initiation: identifying contradictions and conflicts in results to refine research design; and (5) expansion: extending the scope, breadth, and depth of research through diverse methodologies. These fundamental features underscore the practical value of systematic reviews and inform the overall research design. In essence, the systematic review method facilitates a multi-faceted analysis of research questions, thereby enhancing the comprehensiveness and depth of the literature review.

Fig. 1
figure 1

Systematic review framework.

Full size image

Literature filtering

The quality of relevant data in the literature is paramount to this study, particularly for the quantitative interpretation of knowledge domains through bibliometric analysis. As such, we followed the literature collection criteria of Moher, et al.22 (1) contemporaneity and relevance: all selected papers were published between 2014 and 2024, with subsequent manual screening of keywords and abstracts to ensure topical relevance; (2) source journal authority: only papers from journals and conferences with rigorous review processes and high-quality information were included, adhering to widely accepted methodological standards.

This study utilizes the Scopus and Web of Science (WOS) database as the primary search platform. Scopus is currently the world’s largest abstract and citation database, encompassing over 20,000 journals across science, technology, medicine, and social sciences, published and distributed by >5000 publishers23. Its comprehensive and subject-specific nature surpasses other abstract indexing databases, enabling researchers to access a broader spectrum of literature and providing more extensive data support for the research community. Furthermore, to mitigate individual bias and enhance the overall review quality, this paper employs a structured data search process (Fig. 2), aligning with the meta-analysis protocol for systematic reviews24.

Fig. 2
figure 2

Structured data search process.

Full size image

Advanced search by title, abstract and keywords as follows: TITLE-ABS-KEY (virtual technologies) OR (virtual reality) OR (augmented reality) OR (mixed reality) and TITLE-ABS-KEY (culture heritage) OR (architectural heritage). These search terms were primarily derived from previous literature review studies. Although the application of virtual technology in architectural heritage first emerged in 200125, the rapid advancements in computer technology necessitate consideration of methodological timeliness. Consequently, the search was limited to the period 2014-2024, providing a decade of cutting-edge research data sufficient for contemporary analysis.

The initial search yielded 2264 related journal articles. To ensure relevance determination during screening, explicit inclusion and exclusion criteria were applied: (a) INCLUDE studies applying VR/AR/MR technologies to cultural or architectural heritage documentation, conservation, or interpretation; (b) EXCLUDE studies focused solely on non-heritage applications, technical hardware development without heritage context, non-immersive technologies, theoretical papers without empirical application, or non-English publications. Further refinement through examination of titles, abstracts, and keywords led to the exclusion of 1924 papers, and an additional 12 papers were eliminated following full-text review. The final sample size of 328 papers was justified as it represents a focused corpus after rigorous screening, sufficient for robust bibliometric analysis while maintaining analytical depth. Screening was conducted independently by two reviewers with disagreements resolved through consensus.

Bibliometric analysis tools

In contrast to traditional methods, bibliometric approaches employ data mining techniques to link literature concepts, thereby uncovering implicit knowledge relationships within the literature. This study utilizes VOSviewer, an open-source bibliometric analysis software, to construct the bibliometric network. We selected VOSviewer over alternatives (e.g., CiteSpace, Bibliometrix) because it provides clear visualizations for medium-sized datasets and robust algorithms for link strength normalization. It is particularly effective for keyword co-occurrence and bibliographic coupling analysis26. Consequently, a systematic bibliometric analysis of 328 documents was conducted as follows: First, bibliographic coupling analysis was performed to measure reference similarity and potential connections between documents27. Second, keyword co-occurrence analysis was employed to identify core keywords of related articles and to construct the knowledge network28. These analyses provided the foundation for the subsequent classification of research topics in the review section.

Systematic review method

A systematic review inherently requires qualitative categorization of the screened literature. Topic categorization in systematic reviews requires synthesizing multiple perspectives and refining research topics through discussion to achieve consensus29. Given that a structured approach can provide direction for addressing heterogeneous findings and minimize potential subjective bias in qualitative reviews30, it is crucial to note that this systematic review is grounded in bibliometric analysis methods. In conclusion, systematic reviews not only inform future research directions but also provide theoretical support for topic discussion.

Results

Overview analysis

The temporal distribution of the 328 screened articles, as depicted in Fig. 3, reveals a consistent presence of publications on virtual technologies in architectural heritage conservation from 2014 to 2024. A remarkable surge emerged in 2016, coinciding with the landmark release of HTC Vive31, the first Steam-powered virtual reality (VR) system. This breakthrough marked a pivotal transition from stationary, device-centric implementations to more immersive, user-oriented virtual experiences. The subsequent years witnessed sustained high publication activity, culminating in a peak in 2024. This trajectory reflects not only the technological maturation of virtual solutions but also demonstrates the growing global recognition of their transformative potential in architectural heritage preservation.

Fig. 3
figure 3

Statistics of published years of 50 selected papers.

Full size image

The geographical distribution of research activities exhibits notable concentration patterns (Fig. 4a), with Italy leading at 36.7% of total publications, followed by Spain (23.5%) and China (21.1%). Other contributing nations include Portugal, Turkey, Egypt, Brazil, Australia, Iran, and the USA, though with considerably smaller shares. This distribution pattern suggests that Mediterranean countries, particularly Italy and Spain, have established themselves as primary hubs for research in virtual heritage conservation, possibly due to their rich architectural heritage and strong preservation initiatives. Additionally, Disegarecon emerged as the dominant journal with 73 publications, significantly outpacing other venues (Fig. 4b). Buildings and Journal of Cultural Heritage followed with 41 and 39 publications respectively, while the International Journal of Architectural Heritage and International Journal of Architectural Technology and Sustainability contributed 26 and 13 publications respectively.

Fig. 4: Publication analysis.
figure 4

a Geographic distribution of research by country, and (b) Distribution of publications across top 5 journals.

Full size image

Bibliographic coupling and Keyword co-occurrence analysis

Bibliographic coupling analysis helps identify research frontiers in a field by analyzing the similarity between cited and citing articles on a specific topic. This approach measures the likeness of papers based on the number of co-cited references between two papers27. As shown in Fig. 5. node size represents article citation frequency, while the distance between nodes indicates the similarity of references between papers. Different colors denote distinct clusters, with color similarity within a cluster signifying relevance or similarity. Based on this analysis, VOSviewer categorized the threshold-meeting articles into five clusters: Cluster#1: Virtual Reality and Heritage Reconstruction; Cluster#2: Augmented Reality and User Experience; Cluster#3: Information Management; Cluster#4: Heritage Building Information Model (HBIM); Cluster#5: Heritage Education.

Fig. 5
figure 5

Bibliographic coupling network.

Full size image

Keyword co-occurrence analysis is crucial for identifying the development process and characteristics of a research field, as well as elucidating relationships between fields or disciplines28. Figure 6 illustrates the clustering relationships among the keywords. Node size indicates keyword frequency, while the thickness of the connecting lines between nodes represents keyword similarity. The five clusters are distinctly color-coded and can be characterized as follows: Information modeling (Cluster 1): Centers around “BIM” and “HBIM”, representing the core technical foundation in digital heritage documentation. Data acquisition (Cluster 2): Focuses on data acquisition methods, including “laser scanning”, “photogrammetry”, and “point cloud”, highlighting the primary techniques for capturing heritage site information. Immersive technology (Cluster 3): Encompasses “virtual reality”, “augmented reality”, and “3D modeling”, demonstrating the growing importance of immersive technologies and digital visualization in heritage presentation. Digital documentation (Cluster 4): Focuses on “digital twin”, “preservation”, and “documentation”, reflecting the emerging paradigm of comprehensive digital heritage management and preservation strategies. Intelligent management (Cluster 5): Connects “deep learning” and “management”, indicating the increasing role of artificial intelligence and advanced computational methods in heritage conservation management.

Fig. 6
figure 6

Keyword co-occurrence network.

Full size image

Systematic review

Following a comprehensive literature review and discussion, the 328 identified articles were categorized into the following four integrated research topics: (1) technology approaches; (2) data acquisition; (3) application directions; and (4) behavioral experiences. Figure 7 illustrates how the research themes were grouped using bibliographic coupling and keyword co-occurrence analyses, and shows how they are related to the articles that were systematically reviewed. Each research theme for the systematic review was then divided into several sub-themes, reflecting the informative role of the bibliometric analysis.

Fig. 7
figure 7

Systematic review framework.

Full size image

Topic 1: Technology Approach. Heritage BIM (HBIM) applies Building Information Modelling to cultural heritage. It is increasingly used for systematic documentation, knowledge management, and conservation planning of historic structures. HBIM goes beyond traditional 3D modelling. It creates digital twins that combine precise geometry with historical, material, and structural data from techniques such as terrestrial laser scanning (TLS) and photogrammetry32,33. This integrated approach facilitates the creation of centralized information repositories, which are crucial for understanding complex structures, material properties, and past interventions. Studies develop domain-specific ontologies and semantic enrichment to manage non-standard heritage knowledge. They also address how to represent irregular geometries and historical significance in BIM34,35. The HBIM model thus becomes a dynamic platform for collaborative conservation, storing condition assessments and facilitating archival research and future monitoring and maintenance strategies. This has been demonstrated in case studies such as the National Palace of Sintra36 and post-fire recovery of the National Museum of Brazil37. Effective HBIM implementation faces two main challenges: accurate parametric modelling of irregular heritage elements, and integration of diverse data into coherent workflows. A primary research focus lies in developing efficient “Scan-to-HBIM” processes, leveraging point cloud data for creating accurate as-built models that capture deformations and complex geometries not representable by standard BIM libraries. Studies explore semi-automated segmentation techniques, custom parametric object libraries for specific architectural styles (e.g., Manueline, Baroque), and procedural modelling approaches constrained by historical construction rules to enhance modelling efficiency and geometric fidelity38,39,40. Researchers stress moving from idealized models to “as-built HBIM” that records actual deformations. This is essential for structural analysis and planning41,42. Challenges remain in creating standardized, interoperable workflows that link historical research, survey data, semantics, and specialist analysis. They must also manage computational load and suitable LOD and LOI for conservation aims43. Beyond documentation, HBIM is increasingly recognized as a vital tool for promoting sustainable conservation practices, enhancing management efficiency, and fostering the valorization of cultural heritage. It enables sophisticated simulations, such as energy performance analysis and seismic vulnerability assessment, informing sustainable retrofit strategies that balance energy efficiency with heritage integrity, though interoperability challenges persist44,45,46,47. In management, HBIM acts as a central hub for facility management, maintenance scheduling, and stakeholder coordination. It links spatial and asset data within GIS or web platforms to support access and long-term stewardship48,49,50. Crucially, HBIM supports sustainable cultural heritage tourism by enabling the creation of virtual interactive experiences, improving visitor engagement, and aiding in the management of tourism impacts on fragile sites, contributing directly to Sustainable Development Goals (SDG 11.4)51,52. The technology also facilitates virtual restoration, damage assessment, and the exploration of alternative intervention scenarios before physical work begins, minimizing risk and preserving authenticity53.

Virtual Reality (VR) and Augmented Reality (AR) have transformed cognitive accessibility and public engagement with architectural heritage, particularly at sites with physical limitations or partial destruction. At Hadrian’s Villa, Bertacchi and Adembri digitally reconstructed inaccessible structures and shared them online, enabling inclusive virtual visits and offering new perspectives54. Similarly, Zheng et al. highlighted the role of VR in digitally preserving the Red Pagoda in Fuliang County by combining 3D laser scanning, BIM and VR to transform physical heritage into shareable digital formats55. These technologies transcend traditional barriers by offering multisensory experiences. For example, researchers utilized advanced sound models to convey spatial information to visually impaired users, ensuring they could experience the environment more vividly56. Collectively, these approaches democratize access, foster deeper contextual understanding, and preserve intangible heritage values through immersive storytelling.

Beyond accessibility, VR and AR facilitate precision conservation, virtual restoration, and data-driven management of architectural heritage. AR tools integrated with semantically enriched BIM models have been developed to enable on-site restoration workflows, overlaying historical data, material analyses, and conservation guidelines onto physical structures via mobile devices57. This approach aligns with recent VR-based documentation efforts that capture transient craftsmanship through immersive simulations, enriching preservation archives and public awareness58. For degraded or lost elements, technologies enable non-invasive revival: digitally restored polychrome paintings have been projected onto church surfaces using AR59, while VR has been employed to reconstruct the decontextualized Gothic choir of Girona Cathedral, allowing interactive comparison of historical and current states60. Furthermore, AR has supported geometric analysis of historic urban centers, transforming passive tourism into educational experiences that highlight architectural proportions and evolution. For example, it has promoted sustainable cultural tourism at Beijing’s Yonghe Temple61. These innovations demonstrate the capacity of VR and AR to bridge the gap between conservation needs and sustainable tourism, ensuring the longevity of heritage sites while minimizing physical intervention.

Furthermore, the application of deep learning techniques is rapidly transforming the field of architectural heritage conservation, offering novel solutions for documentation, damage assessment, monitoring, and decision support. A significant focus lies in the geometric restitution of damaged or missing architectural components. Karadag pioneered the use of Conditional Generative Adversarial Networks (cGANs) trained on extensive plan datasets (200 Ottoman tombs) to predict large-scale missing elements such as walls, domes, and openings62. Concurrently, ensuring the reliability of Structural Health Monitoring (SHM) data for preventive conservation is critical. Deng et al. addressed the challenge of abnormal SHM data in ancient structures (e.g., a 600 year-old Beijing city wall) by developing a novel deep learning framework utilizing optimized Gated Recurrent Unit (GRU) networks63. The key innovation is bidirectional prediction, using both past and future data points. This improves recovery of strain and crack-width data affected by temperature.

Automated visual inspection for defect detection is another vital area beyond geometry and monitoring. Saravanan and Bhaskar used advanced computer vision, specifically Mask Region-Based Convolutional Neural Networks (Mask R-CNN), which were trained using a large dataset comprising 501 synthetically generated images and real photographs of UNESCO sites in India64. Their system achieved a high level of accuracy (93–94%) and an Intersection over Union (IoU) score of 91–94% in identifying and segmenting degradation such as vegetation growth on stone heritage structures. This offers a scalable alternative to manual inspection. An emerging line of work integrates AI into management. Bienvenido-Huertas et al. implemented a J48 decision tree within an HBIM environment65. They developed and embedded a classification model to automate decision-making regarding intervention levels for specific elements (e.g., tile panels in the Real Alcázar of Seville), streamlining assessments and information management within the digital twin framework.

Topic 2: Theoretical Methods. The theoretical methods underpinning these technologies begin with data collection and virtual reconstruction. Recent advances in data acquisition technologies have transformed the way complex architectural heritage is documented, enabling the high-resolution, non-invasive capture of challenging structures such as domes and inaccessible sites. Portable structured-light 3D sensors enable on-site digitization with sub-millimeter accuracy, which is crucial for unique and fragile objects66. Additionally, UAV photogrammetry has become indispensable for large-scale or hazardous sites. For example, Li et al. developed CU-Recon software using deep learning for efficient single-drone reconstruction, which has significantly enhanced public engagement in conservation efforts such as those involving Hakka Tulou67. Similarly, ultralight drones (weighing <250 g) have proven effective in steep terrain, as demonstrated by Russo et al. at Canossa Castle, where they integrated seamlessly with GNSS and TLS68. Sender et al. used them for Casa Castellote’s Renaissance palace, generating precise data for historical analysis69. Furthermore, mobile LiDAR systems like Zebedee can rapidly map expansive sites (e.g. Peel Island Lazaret), efficiently capturing structural and contextual details70. Even smartphone-integrated LiDAR offers accessible documentation of complex elements, such as the frontispiece of the Saint Francis of Assisi Church71.

The transition from raw data to semantically rich virtual models relies on sophisticated processing algorithms, geometric analysis, and integrated digital workflows. Point cloud optimization is critical for managing massive datasets, Rodríguez-Gonzálvez proposed classifying points using omnivariance metrics, achieving 85% reduction while preserving detail in complex areas like the Niculoso Pisano Portal72. Zhang et al. developed the TCCWS framework, which uses teacher-student models and contrastive learning to accurately segment point clouds with minimal (0.1%) labelled data73. Additionally, the geometric analysis of acquired data enables a historical understanding and a structural assessment. Bianchini established protocols using active and passive sensors and modelling software to decipher dome geometry74, while Costa-Jover, et al. devised TLS-based methods to detect formal anomalies in Gothic vaults at Tortosa Cathedral75. Moreover, integration into HBIM platforms enhances conservation management. Escudero et al. combined TLS, machine learning (RANSAC, K-Means), and colorimetry to create a semantically enriched model for Valencia Cathedral’s Holy Chalice Chapel that links geometric and material data76. Furthermore, simplified reconstruction tools like CU-Recon democratize participation, fostering bottom-up conservation models67. However, challenges persist in automating the interpretation of complex heritage geometries and ensuring seamless interoperability between diverse datasets within conservation-focused digital twins.

Systematic paradigms and research methods form another critical methodological area. The scale and complexity of architectural heritage and historical settlements often preclude their physical exhibition in museums or galleries. As Gülec notes, even if it were possible to relocate an architectural work, its original meaning would be compromised due to the loss of context77. Consequently, it has become crucial to document and display both the original and altered forms of buildings. Hajirasouli, et al.78 developed a digitally integrated framework for the longitudinal observation of cultural and historical sites, which they applied to conserve an architectural heritage colony in Kandan, Iran. Their approach included an interactive virtual reality platform to encourage participation. Kavakli79 proposed a user-centered mobile augmented reality system, the MARS framework, which facilitates rapid development of diverse applications to address project-specific challenges.

An interdisciplinary approach incorporating sociology, psychology and anthropology has been adopted in architectural heritage conservation research from a virtual perspective. Beyond traditional questionnaires and interviews80, researchers have employed experimental methods for data collection. These include controlled variable studies13,81,82 and scenario settings83,84 to investigate participants’ sensory experiences and system refinement. Both qualitative and quantitative methods have been widely used to evaluate virtual technology frameworks81. Ghani, et al.85 evaluated real-time user feedback for three levels of immersive systems through a qualitative experimental approach, measuring the importance of virtual scenarios in user-place experiences. The sense of immediacy in virtual technology is often considered a critical factor in understanding and assessing the effectiveness of virtual environments86.

Application Directions. The application directions of these virtual technologies are diverse. Virtual technologies offer significant advantages for both research scholars and students in architectural heritage education, providing meaningful experiential support and immersive sensory experiences. In their exploration of different levels of learning experiences, Krokos, et al.87 used augmented projective reality to enhance embodied collaborative behaviour in a Spanish bomb shelter heritage experience. This approach fostered students’ emotional engagement and critical thinking skills. In response to the lack of user-friendly virtual platforms in architectural heritage education, Puggioni, et al.13 proposed the ScoolAR virtual platform. This platform enables users to create content independently through accessible features, thereby enhancing engagement with, and recognition of, AR and VR applications in the field. Jacobson88 provides a representative research methodology and design for assessing the impact of virtual heritage education. The results demonstrate that self-centered immersive experiences facilitate learners’ knowledge acquisition and absorption.

Virtual technology effectively preserves heritage sites while mitigating conflicts between conservation and the utilization of cultural tourism resources. Archeoguide is one of the first applications to use AR technology to provide personalized tours of cultural heritage sites. It allows users to experience a virtual world based on computer-generated 3D reconstructions while remaining connected to the real environment89. Tao and Archambault90 developed various personalized mobile guidebooks for outdoor architectural heritage, with practical applications in ancient Olympian sites25. Panou, et al.91 introduced a novel sightseeing experience for the historic site of ancient Chania in Crete, Greece, inspiring visitors to explore and connect with the underlying history of the urban site. Numerous virtual tourism approaches have been demonstrated92,93. For individuals with physical disabilities, virtual technology has become essential for accessible travel. Pérez, et al.94 developed a virtual reality program to provide realistic sensations for wheelchair visitors, addressing accessibility constraints at the Cancho Roano site. To address this research gap, Stoyanova-Doycheva, et al.95 provided a methodology for generating dynamic tour itineraries. Their approach uses ontological and environmental networks as a knowledge base.

Risk management can identify, assess, and analyze the damage to heritage buildings, or develop strategies to reduce it. Lee, et al.96 proposed a new metadata and risk management framework based on the 5WH1 principle (what, when, where, who, why and how). Combining HBIM and VR technology provides monitoring and diagnostic personnel, as well as those identifying building information, with a wealth of shared data, reducing time and effort. This approach enables multiple professionals to collaborate in a virtual environment while providing a comprehensive data foundation for risk management decisions97,98. Traditionally, updating and upgrading architectural heritage knowledge requires significant government financial support. Marra and Fabbrocino49 addressed this survey gap by leveraging social networking information and web mapping platforms, shifting the heritage survey process from government-led to community-engaged approaches.

Topic 4: Behavior Experiences. Preserving and communicating architectural heritage requires a high degree of accessibility and multisensory experiences. Ioannides, et al.99 emphasize the importance of multisensory reconstruction of cultural heritage environments. These systems recreate environments using temperature changes, sound, touch and vision to provide a more immersive experience. The Augmented Perception for Path Recognition for Indoor Assisted Navigation (ARIANNA) framework provides an innovative solution that enables visually impaired or blind people to visit and experience heritage sites independently100. This system uses a smartphone’s camera to detect floor paths and provides feedback to the user through vibration signals, effectively turning the smartphone into a connecting tool between the user and the environment. Additionally, Antlej, et al.101 proposed a mobile augmented reality application containing historical information about Ovid’s life. This application aims to provide a rich, multisensory experience through computer-generated layers containing visual, auditory, and tactile information in a virtual recreation of a classical museum setting.

Gamification mechanics offer some of the most effective ways to convey sensory experiences through digital media102. These mechanics draw people into virtual representations of the real world that can be explored through active participation. Unlike traditional scenario simulations, games enable players to access the outcomes of their actions, offering various approaches to exploring and interacting with cultural heritage settings103. Historical contexts in games can convey diverse experiences and encourage players to think more deeply about space, place, and time, thus bringing architectural heritage to life. For architectural heritage, contemporary relevance often takes precedence over historical information104. Fazio and Turner105 address the problem of “empty rooms” that tourists often encounter when visiting historical buildings through their Skullduggery heritage project. This augmented reality design, inspired by adventure games, recreates the viewing experience, providing visitors with an immersive view of history and a stronger sense of place. Leveraging the potential of game narratives, this approach provides a more engaging and interactive way to experience architectural heritage.

Discussion

Virtual technology has rapidly advanced and is increasingly applied in architectural heritage conservation and communication, establishing itself as a significant research focus. Since 2016, research groups led by Bruno82,106,107, Panou91, and Haydar84 have emerged, driving intersecting and evolving research directions. Keywords such as “virtual reality,” “augmented reality,” and “HBIM” now dominate the field. Concurrently, various fields and disciplines have integrated, including virtual tourism14, architectural heritage education13, and computer vision108. While virtual technology is extensively researched for heritage conservation, the growing scale and impact of complex environments pose both challenges and opportunities for future work.

Despite the potential of HBIM, IoT and AI for integrated adaptive management, our analysis indicates that fragmentation remains a persistent issue. Cluster 3 (Information Management) and Cluster 5 (Heritage Education) reveal technical barriers through bibliographic coupling, as Cluster 1 (Information Modelling) and Cluster 5 (Intelligent Management) do through keyword clusters. HBIM’s semantic richness often prevents it from dynamically integrating real-time sensor data (e.g. from SHM systems using GRU networks) or AI-driven predictive analytics. Crucially, bidirectional interoperability remains a major issue. Most systems operate as static repositories rather than dynamic learning ecosystems. Future frameworks should therefore adopt federated learning architectures to enable AI agents to continuously refine HBIM models using data from drones, LiDAR and visitor interactions. However, this requires significant standardization of heritage ontologies and raises ethical challenges, particularly with regard to data sovereignty at vulnerable heritage sites. Without addressing these issues, the “adaptive” paradigm risks becoming merely a techno-utopian concept.

The rise in VR and AR applications (Keywords: Cluster 3; Bibliographic Clusters 1-2) highlights a core ontological tension: virtual reconstructions increasingly compete with, rather than complement, physical heritage. Projects like Hadrian’s Villa VR or Girona Cathedral’s virtual choir sometimes prioritize immersive spectacle over material authenticity, potentially undermining conservation ethics. Generative AI intensifies this risk, with cGANs that generate missing architectural elements (e.g., Karadag’s study of Ottoman tombs) having the potential to propagate historically inaccurate reconstructions. However, this technology also has significant potential for problem resolution. AI-driven comparative analysis of degradation rates in virtual and physical models could predict material fatigue and detect building damage. We propose a new approach to virtuality based on critical digital heritage principles. These include mandatory metadata disclosing reconstruction uncertainty and AR interfaces that visualize structural vulnerabilities (e.g. overlaying crack propagation simulations on physical structures). This approach would help to reinforce conservation priorities.

User-focused applications (Bibliographic Cluster 2; Systematic Review Topic 4) have been shown to increase engagement. Virtual technology shifts the heritage experience from being centered on the architecture itself to being centered on the user, offering multi-sensory experiences. However, existing research has not sufficiently explored the actual benefits and utility for users. Further empirical investigation is needed to understand user immersion, aesthetic appreciation, well-being and social interactions enabled by the real-time sharing of virtual heritage. Additionally, the complex relationship between the positive effects of virtual technology and its potential to cause adverse user experiences, such as sensory fatigue and environmental isolation, requires deeper study. Future research could utilize neuroadaptive interfaces to dynamically adjust virtual content based on user responses.

This paper conducted a systematic review of the application of virtual technologies in architectural heritage conservation, using a bibliometric analysis of 328 published papers. We identified key articles and hot topics through bibliographic coupling and keyword co-occurrence analysis. Our analysis reveals three key challenges and future research directions for virtual technologies in this field. However, there are several limitations. Firstly, despite the substantial size of the dataset (n = 328), our reliance on English-language resources may result in valuable insights from non-English-speaking regions, such as those in Mediterranean and Asian contexts, being excluded. Secondly, from an application perspective, the bibliometric coupling used captures established rather than emerging trends. This could result in the overlooking of recent developments such as generative AI and LLMs, which present new opportunities for heritage conservation. To address these limitations, future research should first expand the scope to include databases such as CNKI and CumInCAD. It should also incorporate more studies from non-English-speaking regions, particularly those actively engaged in preserving digital architectural heritage. Furthermore, future studies should actively engage in interdisciplinary research and keep pace with trends in the field. They should also collaborate closely with heritage institutions, technology developers and policymakers to establish ethical frameworks that guide the integration of digital and physical heritage.