AI-Enhanced Co-Creation in Industrial Heritage Architecture Tourism: Exploring Authenticity and Well-Being at the Yangpu Cold Storage Facility

Abstract

As urbanization intensifies, the challenge of preserving industrial heritage while fostering authentic intergenerational connections has become increasingly salient. This study investigates how artificial intelligence (AI) and augmented reality (AR) technologies can be applied to enhance authenticity and promote both hedonic and eudaimonic well-being within the context of heritage tourism. Using a facility in Shanghai as a case study, we propose a cultural co-creation mechanism that transforms implicit intergenerational memories into shared cultural resources through digital interaction. The study first evaluates public awareness and participation needs in the context of industrial heritage revitalization. In response, we design an immersive platform that enables visitors of different generations to co-create meaning through historical scene reconstruction, multisensory engagement, and collaborative storytelling. A novel five-sense encoding strategy is introduced to reinterpret the enclosed spatial characteristics of industrial architecture as an experiential form of storytelling. This process fosters a deeper connection to place, contributing to authenticity and well-being. Prototype testing results suggest that this AI-AR-enabled co-creation system supports meaningful cultural attachment, improves authenticity, and facilitates the sustainable transmission of heritage. This research provides a replicable model for integrating digital technology, community participation, and authenticity in the well-being-oriented revitalization of industrial heritage sites.

1. Introduction

Industrial heritage refers to the legacy formed during the development of industrial civilization that holds historical, technological, social, architectural, or scientific value [1]. As direct products of the Industrial Revolution, industrial heritage buildings record technological progress and craftsmanship innovation of specific historical periods [2,3], serving as crucial material evidence for the study of the history of science and technology [4], and embodying irreplaceable historical, cultural, and social values [5]. According to the World Heritage Convention, the fundamental obligation of the State Parties is to “identify, protect, conserve, present, and transmit to future generations”. The conservation and transmission of industrial heritage buildings thus emphasize generational transfer through management and presentation, while safeguarding authenticity and integrity [6].

With accelerated urbanization, conflicts between industrial heritage sites and urban construction land have intensified, placing industrial heritage buildings at risk of demolition or disappearance [7]. However, traditional conservation approaches tend to adopt static protection, maintaining existing conditions without altering heritage features, thereby delaying deterioration. Guided by the cautious principle of “necessary change, minimal change” [8], such static approaches are limited by the rigid spatial patterns of industrial buildings, which hinder the provision of open and multifunctional uses. Furthermore, the rigid requirement of “non-alteration of the original state” for core items [9] often results in superficial renewal and “conservation-oriented idleness” [10]. Interpretive methods relying on panels and object displays fail to foster in-depth understanding and intergenerational transmission [11]. To meet contemporary demands for multifunctional space and cultural participation, there is an urgent need to shift from the traditional “demolition and reconstruction” paradigm toward a more sustainable “conservation and innovation” strategy [12], in which technology empowers the dynamic symbiosis of industrial heritage conservation.

Dynamic symbiosis refers to an integrated approach that combines conservation with adaptive reuse and contemporary urban life—embedding heritage policy into local development processes, ensuring stakeholder participation, and promoting compatible uses, so that heritage and the city evolve together [13,14].

Developing tourism resources based on industrial heritage buildings can reconcile land use conflicts, optimize urban functional layouts, promote tourism development, preserve urban industrial culture [15], and facilitate the renewal and revitalization of industrial heritage buildings [7].

In cultural and tourism studies, authenticity and well-being are regarded as twin core indicators for evaluating the depth of heritage experience. Since MacCannell [16] introduced “staged authenticity,” scholars have proposed multi-dimensional frameworks including objective authenticity, constructive authenticity, and existential authenticity [17,18,19]. Among these, existential authenticity emphasizes self-realization and emotional resonance derived from immersive experiences, and has been widely applied to explain the mechanism through which heritage tourism enhances well-being [20]. While most studies examine different dimensions of authenticity separately, Kolar, building on Wang’s classification, introduced the concept of “object-based authenticity (OBA)” and argued for simultaneous attention to both OBA and “existential authenticity,” in order to balance the preservation of heritage fabric with the visitor’s emotional engagement [19]. Accordingly, this study also discusses authenticity in terms of both OBA and existential authenticity.

The dual-structured concept of well-being originates in psychology: Ryan and Deci distinguished between hedonic well-being, the pursuit of immediate pleasure, and eudaimonic well-being, the pursuit of self-realization and meaningfulness [21]. In tourism, this framework was consolidated by Smith and Diekmann, who summarized that hedonic well-being aligns with “having fun,” while eudaimonic well-being emphasizes “learning and gaining value” [22]. Substantial empirical evidence demonstrates that authenticity significantly enhances well-being. For instance, surveys in China’s World Heritage Sites revealed that perceived authenticity increases subjective well-being via the chain of “place attachment→satisfaction” [23]; Gallou’s study of UK communities further showed that heritage participation simultaneously improves psychological health and social cohesion [24]. In industrial heritage contexts, the integrity of remains and transparency of narratives significantly boost visitor satisfaction, loyalty, and pro-environmental behaviors [15]. Cross-location model verification indicates that existential authenticity acts as a key mediator linking authenticity, well-being, and memorable experiences [25], with its positive effects remaining robust across cultural backgrounds and tourism motivations [26].

Meanwhile, rapid advances in digital technology—particularly Artificial Intelligence (AI) and Augmented Reality (AR)—offer new opportunities for the conservation and revitalization of industrial heritage buildings. AI’s semantic mining and predictive capabilities bring unprecedented efficiency to the correlation and reconstruction of fragmented historical records [27]. AR, with its spatiotemporal layering and immersive interactivity, transcends traditional display limits, transforming “inaccessible and unintelligible” industrial relics into tangible, playable cultural theaters [28]. It enables visitors to “enter history” in first-person mode, creating multisensory perceptual channels [29,30], while museum-based empirical studies further show that AR-guided tours simultaneously enhance authenticity, presence, and satisfaction [31]. However, existing research has not yet systematically examined the interconnections between AI–AR technologies, authenticity, well-being, and the willingness to conserve and transmit heritage. Current practices often focus on single technologies or unidirectional narratives, overlooking the social dimension of co-creation—the cyclical interaction of stakeholders in generating, interpreting, and governing heritage experiences [32,33].

In response to these challenges, this study proposes the following research questions:

• RQ1: How can AI and AR technologies transform intangible intergenerational memory into shared cultural resources, thereby enhancing authenticity and well-being in industrial heritage tourism?
• RQ2: How can AI–AR technologies break the traditional expert-dominated model to construct a multi-stakeholder participatory mechanism?
• RQ3: What key design considerations should be incorporated to foster intergenerational audience interaction and increase interest in and willingness to conserve industrial heritage buildings?

To address these questions, this study selects the Yangpu Cold Storage complex—an industrial heritage site witnessing the transformation of Shanghai’s cold-chain industry and urban defense system—as the case study. Unlike typical industrial heritage sites, the main buildings of Yangpu Cold Storage adopt a beamless slab structural system, creating a vast column-free space of 30 m × 60 m. However, the 4.2 m floor height and fully enclosed “black box” form (without windows and ventilation facilities) limit space utilization to below 20%. Internally, complete industrial production traces remain, while externally, fragmented renewal is evident: a 45 m × 45 m football field and a 20 m wide riverside green belt in the southeast corner provide leisure functions, whereas the 340 m long freight square remains enclosed by walls. This spatial contradiction makes it a typical testbed for the concept of “conservation-oriented renewal” of industrial heritage.

Based on the Yangpu Cold Storage’s features of large-span factory halls and enclosed windowless structures, this study designs and validates an AI–AR-enabled co-creation mechanism and immersive digital platform. Leveraging AI, AR, and five-sense encoding strategies, the platform reconstructs historical scenes, transforming the enclosed spatial features of industrial buildings into experiential advantages, enhancing immersion and interactivity. By emphasizing public participation and multi-stakeholder collaboration, it fosters intergenerational co-narration and cultural co-creation, thereby elevating the authenticity and well-being of Yangpu Cold Storage tourism. Through an integrated process of literature review, user demand research, prototype system development, and mixed-method evaluation, this study aims to fill theoretical and practical gaps in the digital co-creation of industrial heritage, while providing replicable reference pathways for the revitalization of similar industrial heritage sites worldwide.

2. Formative Research

2.1. Pre-Investigation

In order to investigate the public’s expectation of participation in IHA tourism renewal and its perceived effect on the assistance of digital media technology represented by AI and AR, this study adopts a questionnaire to conduct a pre-investigation to collect the public’s opinions on the “authenticity”, “ well-being” and “co-creation mechanism”, and to refine the critical path of system design accordingly.

The questionnaires were distributed through the Questionnaire Star platform from December 2024 to January 2025; 224 questionnaires were collected, and 222 valid samples were retained after excluding invalid data. The overall structure of the questionnaire is divided into four parts, demographic information (Q1–Q5), exposure perception (Q6–Q11), current situation evaluation (Q12–Q19), and future expectation (Q20–Q27), covering the four dimensions of industrial heritage: physical space, cultural narrative, social participation and technological media.

In order to verify the reliability of the data, this study calculated the reliability level of each dimension and the overall questionnaire based on the validly collected questionnaire data using SPSS AU V24.0 online statistical analysis platform. The calculation results show that the overall Cronbach’s alpha coefficient of the questionnaire is 0.897, which is extremely high in reliability. Among them, the technology function dimension has the highest reliability ($\alpha =0.939$), indicating that the questionnaire measures a high degree of consistency in terms of preferences for the application of numerical intelligence technologies such as AI/AR (

Table 1. Questionnaire reliability.

Scope of Analysis	Number of Items	Sample Size	Cronbach’s $\mathit{\alpha }$
Physical Function Dimension	4	222	0.793
Cultural Function Dimension	4	222	0.800
Technical Function Dimension	12	222	0.939
Overall Questionnaire	34	222	0.897

).

Among the respondents, 50.45% are young people aged 18–30 years, and 77.47% hold a bachelor’s degree or above, indicating a strong motivation for technology exposure and cultural participation. Nearly 80% have visited industrial heritage sites in the past one year; their most common motives are participation in exhibitions or activities (65.32%) and history learning (64.86%). However, participation frequency is polarised, suggesting that existing content lacks sustained attraction (see Figure 1).

In terms of the public’s perception of the current situation, the survey found that there is a significant fault line in the perception of authenticity. respondents over 51 years old recognize the authenticity and emotional value of architectural history displays, while the youth group shows a higher preference for interactivity and scenario-based narratives, revealing a disconnect in the mode of intergenerational cultural experience.

In terms of technological participation, there are both expectations and barriers. Although more than 68% of respondents would like to participate in the renewal of industrial heritage through AR/VR, more than half of them think that it is “complicated to operate the technology” (rating 3.43/5), and especially those with higher education level have higher standards for the system experience (Figure 2).

In terms of sensory interaction, respondents demonstrated a strong demand for multi-sensory technologies (visual, tactile, and auditory), which were unanimously considered to be helpful in enhancing emotional immersion and depth of understanding, especially haptic experience (3.50/5) and AI intelligent guides (3.47/5), which were regarded as an important means to improve the sluggishness of traditional displays (Figure 3).

In terms of engagement mechanisms, the survey data reveals a perception that the mechanisms that currently exist are weak and one-dimensional. Although more than 52% of the respondents expressed their desire to participate in the renewal of the heritage, the rating of “channels of participation” was only 2.62/5. The social mechanisms are still at the stage of symbolic mobilization and lack the substantive decision-making power of the public.

In conclusion, this phase of the research reveals that the public has a real willingness to participate in the renewal of industrial heritage and a technical interest in it, but their experience of co-creation is limited by the operational complexity of the current platform, the closed nature of the participation mechanism, and the disconnection of intergenerational narratives. Therefore, the follow-up system needs to build an experience mechanism with “perceivability”, “intergenerational co-creation” and “ease of operation” as the core.

2.2. Design Considerations

Combining the public needs and participation bottlenecks revealed in Section 3.1, the study proposes a design strategy for the AI-AR co-creation mechanism for the Yangpu Cold Storage scenario, and builds a co-creation mechanism that integrates sensory encoding, intergenerational collaboration and emotional connection accordingly.

With the goal of enhancing the perception of authenticity and promoting well-being, this system proposes the following four key design considerations (D1–D4):

D1: Interaction Mechanisms for Multi-Generational Adaptation

In response to the preference of youth groups for high interactivity and the reliance of elderly groups on emotional narratives, the platform needs to design task paths using graphic juxtaposition, voice navigation and tactile prompts in a synergistic manner to ensure that different generations can experience them without barriers.

D2: Five-Sense Embedded Spatial Coding

This involves using AR (augmented reality) to reconstruct the historical production and life scenes of Yangpu Cold Storage, forming a “sensory memory package” through the synergistic feedback of vision, hearing and touch to enhance the emotional embedding of the user and the space, and reshaping the “authenticity in immersion”.

D3: Personalized AI Cultural Guidance Module

We build a recommendation engine based on user behavior and emotional response, matching co-creation tasks, guided paths and cultural materials to realize the transformation from “user-viewed content” to “user-created narratives”.

D4: Public-oriented Co-creation Cycle Model

Based on Nonaka and Takeuchi’s SECI Knowledge Cycle Model [34], we propose the “AI-AR-Driven Cultural Memory Co-creation Mechanism”, which is divided into the following categories:

S (Socialization): Reconstructing historical scenes through AR interaction and guiding users to upload memories, images and narratives; E (Externalization): the platform structures user input into tags, timelines and spatial events through AI semantic analysis; C (Combination): the system automatically builds multimodal visual archives and immersive theaters for users to access and recreate; I (Internalization): the user transforms the experience of participation into an emotional identity and cultural connection and participates in the next round of content feedback and re-editing.

Through the continuous cycle of this SECI spiral structure, the system realizes a closed loop from “narrative co-creation” to “emotional recognition” to “cultural reproduction”. This mechanism not only solves the problems of lack of updating of cultural content, single display method and lack of sense of belonging, but also provides a theoretical model with generalizable value for the digital renewal of industrial heritage (Figure 4).

3. AI-AR-Driven Industrial Heritage Co-Creation: Mechanism Design and System Realization

3.1. Theoretical Framework and Site-Specific Context

The methodological workflow of this study is shown in Figure 5: design considerations (D1–D4) are derived from formative research; within the single-building boundary we implement a five-sense-encoded AI–AR system; the experience is delivered and evaluated using mixed methods, yielding the results and future directions.

The integration of artificial intelligence and augmented reality technologies in heritage tourism represents a paradigm shift from static preservation to dynamic co-creation [35,36]. Recent applications demonstrate AI’s capacity for semantic extraction from historical materials and AR’s ability to create spatial–temporal overlays, forming what the authors of [31] describe as a “data acquisition–semantic parsing–immersive display” technical chain.

The Yangpu Cold Storage facility, while comprising dual warehouse structures connected by loading platforms, presents a focused research site for our AI-AR implementation. This study specifically targets the main warehouse building as the primary spatial unit for digital intervention, acknowledging but not attempting to address the broader composite site configuration. The selected building—with its 8-floor/32.27 m vertical dimension and completely windowless configuration—provides a controlled environment for AR deployment. Importantly, our system design assumes users remain within a single spatial volume, with the AI-AR platform enabling location-specific content activation based on user-selected positions within this defined space. This approach eliminates the technical complexities of inter-building transitions and multi-structure navigation, allowing for deeper engagement with the industrial heritage narrative within a bounded architectural context (Figure 6).

The facility’s distinctive architectural characteristics—seamless precast concrete façade with 5 cm precision joints and the “black box” configuration—originally constraining traditional interpretation methods, paradoxically provide optimal conditions for contained AR experiences. The controlled lighting environment eliminates variability issues common in outdoor AR applications [29], while the 30 m vertical space enables multi-layered narrative construction without requiring physical movement between discrete structures.

3.2. Three-Dimensional Integration Architecture

Our proposed AI-AR co-creation mechanism centers on the WeCool platform, implementing a “three-dimensional integration structure” that synthesizes physical space and digital platforms (Figure 7).

The physical-layer transformation occurs within the selected warehouse structure through AI-enhanced spatial analysis and AR-enabled recontextualization. We utilize photogrammetry combined with LiDAR scanning to generate precise 3D spatial models of the interior volume. The system employs OpenAI’s GPT-4 API for natural language processing of archival documents, extracting spatial–temporal relationships specific to this building’s operational history. Computer vision models, specifically CLIP [37], enable multimodal alignment between historical photographs and current spatial configurations within the defined architectural boundary.

The digital platform layer, built upon Unity 3D engine with ARCore/ARKit SDKs, serves as the primary mediation interface, with spatial anchoring calibrated specifically to the single building’s coordinate system. Users initiate experiences by selecting their position within the warehouse through the platform interface, triggering location-specific AR overlays without requiring physical navigation between multiple structures. This design choice addresses practical constraints of heritage site visits while maintaining immersive engagement within a coherent spatial narrative (Figure 8).

3.3. Sensory Encoding Strategy and Experiential Design

Addressing the selected warehouse’s inherent sensory deprivation—characterized by absence of natural light, ambient sound, and thermal variation—we implement a “five-sense encoding strategy” that reconstructs the facility’s operational atmosphere through multimodal AR overlays. This approach extends beyond visual augmentation to encompass comprehensive sensory reconstruction within the bounded space, following principles established in immersive heritage experiences [38].

Visual reconstruction employs AI-enhanced image synthesis using fine-tuned Stable Diffusion models trained on archival photographs specific to the studied building. The system generates historically accurate 360-degree holographic projections calibrated to the warehouse’s specific temporal periods (e.g., 1966 construction, 1979 expansion, 1985 peak operations), with users able to select temporal layers through interface controls rather than physical movement. Acoustic simulation utilizes spatial audio processing through Web Audio API, recreating industrial soundscapes including ammonia compressor operations and conveyor mechanisms positioned according to documented machinery layouts within this specific structure.

Haptic feedback mechanisms leverage mobile device vibration APIs synchronized with visual machinery animations, while temperature perception is induced through chromatic manipulation—cool-toned lighting (6500K) for refrigerated zones versus warm tones (3000K) for mechanical areas—exploiting cross-modal sensory associations.

3.4. Historical Narrative Path and Intergenerational Co-Creation Mechanism Design

The WeCool platform operationalizes co-creation through four integrated functional modules that facilitate intergenerational knowledge transfer within the spatially bounded context (Figure 9). The AR navigation module serves as the primary experiential interface, enabling users to trigger historical scene overlays through mobile device scanning at their selected position within the warehouse (Figure 10). The system presents temporal navigation options through an interactive timeline interface, allowing users to select specific years and witness corresponding operational states without physical movement. This temporal layering approach transforms the static space into what we term a “chronological palimpsest,” where multiple historical moments coexist within the same spatial coordinates.

The Memory Workshop module implements participatory archiving through structured content collection and AI-assisted processing (Figure 11). Users contribute memories through multimodal inputs—voice recordings, photographs, and text descriptions—which undergo semantic analysis using BERT models optimized for Chinese language processing. The system’s interface design emphasizes intuitive interaction through voice-wave animations synchronized with text transcription, providing immediate feedback that enhances user engagement. Each memory contribution generates a structured “memory card” containing temporal markers, spatial references, and thematic tags, with contributor attribution ensuring transparent provenance. The AI assistance function suggests complementary details based on extracted keywords, helping users elaborate their narratives while maintaining authentic personal voice.

The Co-governance Space module facilitates collaborative decision-making through virtual prototyping within the defined warehouse space (Figure 12). Users interact with 3D spatial proposals through a virtual sandbox interface, manipulating design elements such as rooftop landmarks and interior configurations. The panoramic viewing mode enables 360-degree examination of proposals, with gesture-based controls maintaining accessibility for users with varying technical literacy. Voting mechanisms aggregate preferences across demographic groups, ensuring inclusive representation in heritage interpretation decisions.

The social interaction module implements culturally contextualized gamification through the “Cool Honor” badge system, comprising twelve unique achievements that metaphorically represent the cold storage’s transformation narrative (Figure 12). Each badge features dual-sided design—the front displaying themed illustrations with ice-melting visual effects upon achievement, the reverse presenting historical photographs or future plans. The progressive “thawing” animation symbolizes user contributions gradually warming the frozen industrial memory, creating emotional resonance that transcends mere point accumulation. Badge display varies contextually within user profiles and social feeds, fostering community recognition while avoiding trivializing heritage content.

4. User Study and Evaluation

We conducted a mixed-methods empirical investigation to evaluate whether the AI-AR co-creation mechanism enhances authenticity and well-being perceptions in industrial heritage tourism contexts, and to assess its efficacy in fostering intergenerational engagement and heritage preservation intentions. This evaluation specifically examines the theoretical model’s capacity to address the core challenges identified in industrial heritage revitalization: cultural memory discontinuity, limited public participation, and constrained spatial functionality.

The study recruited 208 participants (ages 18–60, representing 23 occupational sectors and 4 educational strata) to ensure demographic and generational diversity, with 10 participants possessing direct Cold Storage facility experience selected for in-depth qualitative interviews, including 2 former factory workers (ages 45–55), 3 design professionals, 2 architectural researchers, and 3 community members. This purposive sampling strategy ensured comprehensive representation across stakeholder groups while maintaining methodological rigor in capturing diverse experiential perspectives.

The quantitative investigation employed the WeCool prototype platform for experiential engagement, with data collection administered through standardized online questionnaires. The research protocol first provided comprehensive functional descriptions of the WeCool platform, followed by sequential presentation of interaction prototypes encompassing the multi-level interface architecture and four primary functional modules. This structured exposure facilitated participants’ conceptual understanding of the platform’s AI-AR capabilities and co-creation mechanisms before assessment. The validated measurement framework assessed four latent constructs through fourteen items utilizing a five-point Likert scale (1 = strongly disagree, 5 = strongly agree). The AI-AR co-creation mechanism construct (AC) comprised five items measuring immersion enhancement through AR technology in enclosed industrial spaces, perceived usefulness of AI-driven narrative structuring, intergenerational collaborative capability, technical accessibility across age cohorts, and content integration quality [29,39]. The authenticity construct (AU) incorporated three items assessing object-based authenticity of digitally reconstructed scenes, existential self-oriented emotional connection, and interpersonal authenticity enhancement [18,19]. The well-being construct (WB) evaluated hedonic dimensions through positive affect generation, eudaimonic personal growth through historical knowledge acquisition, and relational meaningfulness in intergenerational connections [21]. The preservation intention construct (PI) measured conservation awareness enhancement, destination recommendation likelihood, and future participation willingness [40].

The ten participants selected for qualitative investigation received structured task protocols following their quantitative assessment participation. The experimental protocol comprised two phases: an introductory phase requiring age-display mode configuration and initial system familiarization, followed by an in-depth engagement phase involving virtual sandbox landmark evaluation and memory fragment submission. These structured tasks established standardized criteria for assessing functional familiarity and interaction fluency while enabling comprehensive exploration of the WeCool platform’s capabilities. Following task completion, participants engaged in semi-structured interviews addressing subjective experiential dimensions and system optimization recommendations.

The quantitative analysis yielded 208 valid responses with descriptive statistics revealing positive evaluations across all constructs: AC = 3.74 ± 1.23, AU = 3.71 ± 1.29, WB = 3.75 ± 1.21, and PI = 3.80 ± 1.21 (Figure 13). The one-sample t tests against the midpoint of the Likert scale were significant ($\left|t\right|\ge 13.73$, df = 623–1039, p < 0.001), indicating substantial positive perceptions of system immersion, authenticity enhancement, well-being generation, and preservation motivation. Scale reliability analysis revealed Cronbach’s $\alpha$ ranged from 0.604 to 0.708, with marginally lower values for WB and PI constructs remaining within acceptable parameters. Composite reliability (${\rho }_c$) exceeded 0.79 across all constructs, confirming robust internal consistency. The AC construct demonstrated marginally suboptimal average variance extracted (AVE = 0.461) relative to the 0.50 threshold, attributable to its multidimensional scope encompassing AR immersion, AI narrative integration, and multigenerational usability. The remaining constructs exhibited adequate convergent validity with AVE values between 0.56 and 0.63.

Structural equation modeling utilizing SmartPLS V4.1.0.9 revealed significant positive path coefficients: AC significantly predicted ($\beta =0.693$, p < 0.001) WB ($\beta =0.409$, p < 0.001) and PI ($\beta =0.720$, p < 0.001), while AU demonstrated significant positive effects on WB ($\beta =0.440$, p < 0.001). The explained variance proportions ($R^2$ = 0.48 for AU, $R^2$ = 0.61 for WB, and $R^2$ = 0.52 for PI) indicate that the AI-AR co-creation mechanism accounts for 48–61% of outcome variance, demonstrating substantial predictive validity for the theoretical model.

Bootstrap analyses confirmed the mediating role of authenticity in the relationship between co-creation mechanisms and well-being outcomes. The AI-AR co-creation construct significantly predicted authenticity (t = 18.07, p < 0.001), well-being (t = 6.80, p < 0.001), and preservation intention (t = 20.49, p < 0.001), with authenticity significantly predicting well-being (t = 7.37, p < 0.001). These findings validate both direct effects of the AI-AR co-creation mechanism on outcome variables and indirect effects through authenticity mediation, providing robust empirical support for the proposed theoretical framework (Figure 14).

Qualitative analysis revealed consistent positive evaluations of interface design, functional architecture, and immersive qualities, with particular emphasis on visual coherence with Cold Storage historical narratives. Eight participants explicitly recognized AI technology integration as innovative adaptation to digital-era heritage tourism requirements, conducive to active user engagement. Five participants specifically highlighted the novelty and engagement value of AR-enabled memory reconstruction functionality, reflecting the distinctive contribution of multisensory encoding strategies in industrial heritage revitalization. Unanimous agreement emerged regarding authenticity enhancement through AR-mediated historical scene reconstruction, with majority consensus that the AI-AR co-creation mechanism elevated well-being dimensions during industrial heritage engagement. Participants reported simultaneous affective satisfaction and knowledge enrichment, expressing positive orientations toward intergenerational social connection establishment.

The qualitative findings also identified implementation challenges, including concerns regarding complete socialization utility, insufficient expert response prompting clarity, and uncertainty regarding practical expert contribution utilization. These observations provide critical direction for subsequent system optimization. Despite identified limitations, the application achieved primary objectives with mean scores exceeding neutral midpoints across all measurement dimensions. The evaluation confirms theoretical model feasibility while validating core system value in participation threshold reduction and heritage awareness enhancement, though specific functional elements require iterative refinement. The comprehensive evaluation yields actionable insights for platform optimization, ensuring enhanced utility in future industrial heritage tourism applications.

5. Discussion and Future Work

Based on the knowledge transformation model, this study constructs the “AI-AR-driven co-creation mechanism of industrial heritage cultural memory”, which reveals the dynamic transformation mechanism of cultural memory from individual experience to collective assets, and provides a new theoretical framework for the revitalization of industrial heritage. Taking the case of Yangpu Cold Storage Renewal as an example, this study realizes the immersive reproduction of historical scenes, digital translation of intergenerational memories, and synergistic governance of multi-party subjects through the three-dimensional linkage mechanism of “Physical Space–Digital Platform–Social Network”. Meanwhile, the theory verifies the positive link of “AI-AR co-creation mechanism (AC) → Authenticity (AU) → well-being (WB) → Willingness to preserve/inherit (PI)” through structural equation modeling: AC has a significant positive effect on AU, WB, and PI, and AU significantly enhances WB, and the model explains 48–61% of the variance of AI-AR. This finding suggests that technological immersion not only reconstructs object authenticity, but also activates “existential authenticity” to form Collaborative Authenticity, which ultimately motivates the public to transform emotional identity into protective behavior.

The innovations of the study are mainly reflected in two aspects:

The AI-AR “five-sense coding” technology adaptation method is proposed, which transforms the spatial closure of industrial buildings into immersive experience advantages through multi-sensory interaction design, and solves the dilemma of renewing the physical environments of industrial heritage conservation buildings.

A dynamic path of cultural memory transformation and co-creation is constructed, using AI-AR as a medium to realize the digital collection of personal historical memory, the visual translation of cultural resources and the dynamic generation of public cultural products, forming a sustainable closed loop of cultural heritage revitalization. The study verifies the technology-enabled co-creation mechanism and proves that it effectively solves the problems of cultural memory disconnection, inefficient space utilization and public participation in industrial heritage protection, providing a replicable practice sample for future sustainable heritage renewal.

As an emerging research focus, although the expected results have been achieved, there are still many aspects that deserve further in-depth exploration and improvement in the future.

-

This study has been validated mainly through the case of Yangpu Cold Storage, which may not adequately cover the characteristics of all industrial heritage types and thus has certain limitations. Future research should expand to a wider range of industrial heritage types and explore the adaptability and differences of the AI-AR technology in light of the needs of different geographical and cultural backgrounds.

-

In terms of scale design, the AC scale measures the three elements of “technology immersion, narrative integration, and collaborative ease of use” at the same time, which has a large semantic span, resulting in a low average variance extraction (AVE). In the future, AC can be divided into sub-dimensions or objective behavioral indicators (length of stay, number of UGCs) can be added to improve the reliability and validity and achieve triangular validation.

-

The sample size of the current prototype test is relatively small, making it currently unable to fully reflect the needs and feedback of different groups on the intervention of the AI-AR co-creation mechanism. In the future, the scale of the test should be expanded to cover more diversified user groups, so as to provide more representative data for the application of the AI-AR co-creation mechanism in industrial heritage tourism and conservation.

6. Conclusions

This investigation empirically validates the efficacy of AI-AR-enhanced co-creation mechanisms in addressing fundamental challenges of industrial heritage revitalization at the Yangpu Cold Storage facility. Through mixed-methods evaluation combining structural equation modeling with phenomenological analysis, the study demonstrates that digital technologies successfully transform intangible intergenerational memories into shared cultural resources while democratizing heritage interpretation. The SECI-based cultural memory co-creation mechanism operationalizes knowledge transformation through four integrated stages, with empirical evidence revealing significant enhancement of both authenticity and well-being dimensions. Specifically, the AI-AR platform generates dual authenticity outcomes: strengthening object-based authenticity through historically accurate scene reconstruction (M = 3.71) while simultaneously fostering existential authenticity through emotional connection and intergenerational interaction. The structural equation model confirms that enhanced authenticity mediates the relationship between technological engagement and well-being (b = 0.440, p < 0.001), with participants reporting both hedonic satisfaction and eudaimonic growth through knowledge acquisition and meaningful social connections.

The research identifies essential design parameters that reconcile technological innovation with heritage preservation imperatives. The five-sense encoding strategy transforms architectural constraints into experiential advantages, while the Memory Workshop’s AI-assisted content structuring and Co-governance Space’s participatory mechanisms facilitate inclusive heritage co-creation across generational cohorts. These findings challenge prevailing assumptions that digital interventions compromise heritage authenticity, demonstrating instead that thoughtfully implemented AI-AR technologies deepen authentic engagement by enabling multisensory historical immersion and personal meaning-making. The validated path model (AC → AU → WB → PI) establishes that technologically mediated authenticity experiences translate into enhanced well-being and subsequent preservation intentions ($R^2$ = 0.52), suggesting sustainable behavioral impact beyond immediate experiential outcomes.

This study contributes validated measurement instruments for assessing AI-AR’s impact on heritage tourism experiences and establishes generalizable frameworks for digital heritage revitalization. The empirical confirmation that AI-AR co-creation mechanisms simultaneously enhance authenticity perceptions and well-being outcomes while fostering preservation intentions provides critical evidence for technology-enabled heritage futures. As industrial heritage sites globally confront urban development pressures and cultural discontinuity, this research offers empirically grounded pathways that preserve historical significance through innovative digital engagement, positioning AI-AR technologies as fundamental enablers of authentic cultural memory transformation and sustained community heritage stewardship.