The Intellectual Structure and Thematic Evolution of Trust in AI Research: A BERTopic Analysis

You’re listening to “The Intellectual Structure and Thematic Evolution of Trust in AI Research: A BERTopic Analysis,” by M.-K. Kim, J. Hyun, and J. Lee. Published in 2026.

Abstract.

As AI is increasingly used across social and organizational domains, trust in AI has emerged as an important research topic. This study analyzes 605 articles from 2016 to 2025 to examine the intellectual structure and thematic evolution of the field. Bibliometric and BERTopic analyses were used to identify transitions before and after generative AI (GenAI). The results show that AI trust research expanded rapidly after GenAI, with notable growth in publications and citations. International collaboration analysis reveals a hub-oriented structure in which the United Kingdom holds the most extensive collaborative ties and the United States plays a prominent bridging role across country groups, while Asian countries have grown more visible in recent years.

Topic modeling identified 13 topics across six research areas: AI governance and trustworthiness, organizational and service contexts, XAI and trust, medical AI and trust, AI trust in education, and AI trust in media. Prior to GenAI, research was concentrated around governance, human-AI collaboration, and XAI. Since then, scholarly attention has broadened toward education, media, and the public sector. These findings suggest that AI trust research has expanded toward broader social application domains in the post-GenAI period.

Introduction.

As AI technology is increasingly applied across various social domains, including healthcare, finance, and public adminis-tration, concerns about risks such as algorithmic errors, bias, and privacy violations have also grown. In this con-text, trust in AI carries significant importance: at the technical level, it relates to ensuring the accuracy and transparency of AI systems; at the level of human-AI interaction, it functions as a psychological basis through which users accept uncer-tainty and adopt the technology; and at the societal level, it serves as an institutional foundation for the safe deploy-ment of AI through regulation and ethical standards. Accordingly, trust in AI has emerged as a central topic in AI

The associate editor coordinating the review of this manuscript and approving it for publication was Majdi Mansouri.

research, and the volume of related studies has continued to grow.

The release of ChatGPT in late 2022 marked a turning point, as generative AI (GenAI) rapidly gained widespread use and brought new dimensions to AI trust research. While GenAI is capable of performing a wide range of tasks, includ-ing natural language generation, code writing, and image synthesis, it also introduces new risks such as hallucination, biased outputs, and privacy concerns. These char-acteristics have intensified discussions around the reliability and accountability of AI systems, and it has been suggested that the emergence of GenAI may have brought changes not only in the volume of AI trust research but also in the thematic scope and areas of scholarly interest.

As the body of AI trust research has expanded, review studies aimed at mapping the research landscape and identi-fying knowledge structures have also been conducted. These studies fall broadly into two types. The first comprises works that provide comprehensive overviews of AI trust concepts and research trends, examining the components and con-ditions of trust formation with a focus on explainability, transparency, fairness, and accountability. The second includes studies that analyze AI trust within spe-cific application domains such as healthcare, education, and autonomous systems, identifying domain-specific factors and contextual conditions that influence trust formation. These studies have made important contributions to clarifying the conceptual foundations of AI trust and under-standing research trends across different fields.

However, existing studies have tended to focus on par-ticular technical factors or application domains, or have emphasized bibliometric analysis of research productivity and collaboration structures. Despite the rapid growth of AI trust research following the emergence of GenAI, little is known about how this technological shift has affected the thematic structure and development of the field. There is therefore a need to examine how the major research topics in AI trust have changed before and after the emergence of GenAI.

Against this background, this study analyzes 605 peer-reviewed articles on AI trust published between 2016 and 2025, combining bibliometric analysis with BERTopic-based topic modeling to investigate the intellectual structure and thematic evolution of the field. By comparing the periods before (2016∼2022) and after (2023∼2025) the emergence of GenAI, this study examines how the technological shift has influenced the thematic structure and research trends in AI trust scholarship. The findings are intended to provide a comprehensive account of how AI trust research has devel-oped and to offer implications for future research directions.

II. RESEARCH BACKGROUND.

A. THE CHARACTERISTICS, AND SIGNIFICANCE OF TRUST IN AI

Trust is generally defined as a psychological state in which one is willing to accept vulnerability to another’s actions, based on the expectation that the other party will act in one’s interest. In the context of AI, trust refers to users’ willingness to accept the suggestions or decisions of a system, share tasks with it, and support its use. Prior research has described trust in AI as a multidimensional concept com-prising ability, benevolence, and integrity, and more recently, predictability has also been identified as an important dimen-sion. It is also important to distinguish between trust and trustworthiness. Trust refers to the subjective psy-chological state of the user, whereas trustworthiness refers to the objective properties of the system, such as accuracy, safety, and ethical alignment.

As a result, users may distrust a system with high trustworthiness, while excessive trust may be placed in a system with low trustworthiness.

Trust in AI has several characteristics that distinguish it from interpersonal trust or trust in conventional technologies. First, trust in AI is not a fixed state but a dynamic process that evolves continuously through actual use experience, begin-ning with initial interactions. Second, trust in AI is multidimensional in nature, shaped by both technical indi-cators such as accuracy, safety, and robustness, and ethical indicators such as fairness, transparency, and accountabil-ity. Third, the way trust forms may vary across cultural contexts, with reported differences between cultures that prioritize transparency and explanation and those that place greater weight on system capability and social pres-ence.

The significance of trust in AI is evident across tech-nical, human-AI relational, and institutional dimensions. At the technical level, trust functions as a psychological basis that allows users to engage with AI systems even with-out fully understanding their internal operations, given the non-transparent character of many AI systems. In response to this challenge, explainability and transparency have been emphasized as key means of supporting trust for-mation, with system designs that help users understand how decisions are made gaining increasing attention. At the level of human-AI relations, insufficient trust can lead to the underuse of capable systems, while excessive trust may result in uncritical reliance on flawed ones.

Trust calibration, the alignment between a system’s actual performance and the user’s level of trust, is therefore consid-ered a prerequisite for safe and effective AI use. As AI increasingly functions as a collaborative agent, human factors such as technical literacy, risk perception, and anthropomor-phism have also been recognized as important influences on trust. At the institutional level, regulatory frame-works, ethical standards, and data protection policies serve as the foundation for societal trust in AI. In high-stakes domains such as healthcare and public administration, institutional trust plays a particularly important role in addressing risks such as algorithmic bias, unclear accountability, and privacy violations.

B. A REVIEW OF BIBLIOMETRIC STUDIES ON AI TRUST

As AI technology has spread across various social and indus-trial domains, trust in AI systems has become an important research topic. In response to this growth, bibliometric anal-yses and systematic literature reviews aimed at mapping research trends and identifying knowledge structures have also been conducted. For this study, a search of Scopus for bibliometric analyses and literature reviews on AI trust yielded a total of 48 articles. These studies were found to have been conducted predominantly in recent years, indicating that as AI trust research has grown in volume, the need to organize its trends in a structured way has also increased. The studies can be broadly divided into two types: those that address the concept and structure of AI trust, and those that analyze AI trust within specific application domains.

Several studies have organized the conceptual structure and research trends of AI trust. These works examine the conditions for trust formation with a focus on explainability, transparency, fairness, and accountability. Studies providing overviews of the conceptual structure and research trends in AI trust have been reported,, and research discussing the importance of trust formation in the design and deploy-ment of AI systems has also been presented. Studies analyzing the role of trust in AI-based decision-making environments and research examining the conditions for trustworthy AI in organizational and societal contexts, have also been reported. These studies contribute to clarifying the key concepts and components of AI trust and identifying the factors that influence trust formation.

Review studies analyzing AI trust within specific applica-tion domains have also been conducted. In healthcare, as the use of AI-based diagnostic systems and clinical decision support systems has expanded, studies examining trust for-mation factors in medical settings have been reported. Research analyzing the effects of explainability and transparency on user trust, and studies exam-ining AI trust issues in automated and autonomous system environments have also been presented. In education, research analyzing trust issues in GenAI-based learning envi-ronments has been reported, and systematic literature reviews examining user trust formation across various service contexts have been conducted.

These existing studies have contributed to organizing con-ceptual discussions on AI trust and research trends, and to analyzing scholarly productivity and collaboration structures through bibliometric methods. However, they tend to focus on particular technical factors or application domains, and studies that systematically examine how the rapid growth of AI trust research following the emergence of GenAI has affected the thematic structure and development of the field remain limited. There is therefore a need to analyze how the major research topics in AI trust have changed before and after the emergence of GenAI.

III. RESEARCH METHODS.

Bibliographic data were collected from the Scopus database on January 3, 2026. Scopus is one of the most widely used academic databases for bibliometric analysis, as it comprehensively indexes peer-reviewed journal articles and offers broad international coverage. The search was conducted using the title search field with the query (‘‘artifi-cial intelligence’’ OR ‘‘AI’’ OR GenAI OR ChatGPT) AND ‘‘trust∗’’, limited to journal articles written in English pub-lished between 2016 and 2025. The starting year of 2016 was selected on the grounds that this period is widely regarded as a turning point at which advances in deep learning drew broad public and scholarly attention to AI capabilities and their societal implications, thereby stimulating research interest in AI trust.

The initial search yielded 688 articles, which were then assessed for eligibility through a review of titles and abstracts. This review was conducted by the authors together with one professor and one AI specialist affiliated with a government-funded research institute. Articles were excluded if they did not directly address trust in AI as a primary research theme, for instance, if AI was mentioned only as a peripheral tool or if the concept of trust appeared only incidentally. Where reviewers disagreed on eligibility, a final decision was reached through discussion and consen-sus. Following this process, 83 articles were excluded, and 605 articles were retained for final analysis. The analysis period was divided into two sub-periods based on the release of ChatGPT in late 2022: before the emergence of GenAI (2016 to 2022) and after (2023 to 2025).

This division was intended to allow a comparative examination of how the spread of GenAI may have influenced the thematic structure of AI trust research.

Bibliometric analysis quantitatively examines scholarly lit-erature to map the knowledge structure and developmental trends of a research field. In this study, it was applied to annual publication and citation trends, productivity indicators for core journals and countries, and international collabora-tion networks. Prior to analysis, the raw data exported from Scopus underwent preprocessing to improve data quality. This included standardizing country names, removing dupli-cate records, and verifying author affiliation information to ensure consistency across the dataset. Productivity indicators included total publication count, total citation count, Citation per paper (CPP), h-index, g-index, and m-index. CPP represents the average number of citations per article and reflects citation efficiency adjusted for publication volume.

The h-index simultaneously captures publication productivity and citation impact, the g-index is more sensitive to the influence of highly cited articles, and the m-index reflects the rate of growth in scholarly influence by accounting for years of activity.

To identify core journals and countries, an IQR-based robust z-score criterion was applied. Unlike conventional z-scores, which are sensitive to extreme values, the robust z-score is based on the median and IQR, reducing the distorting effect of skewed citation distributions. Journals and countries that fell within the top 20 rankings in both publication count and citation count under this criterion were considered core. Growth rates between the pre- and post-GenAI periods were calculated using a log-difference approach, specifically ln(post-period + 1) minus ln(pre-period + 1), and were combined with total publication or citation counts to construct portfolio matrices.

Each matrix was divided into four quadrants using the mean values of each axis as boundaries, distinguishing Leading (high volume, high growth), Established (high volume, low growth), Rising (low volume, high growth), and Peripheral (low volume, low growth).

International collaboration was examined through co-authorship network analysis. The network included only countries appearing at least twice and connected through at least two co-authored publications, so as to focus on struc-turally stable collaborative ties. Community detection was performed using the Louvain clustering algorithm proposed by Blondel et al., which identifies communities by iteratively partitioning the network to maximize modular-ity. This algorithm is widely used in scholarly collaboration network analysis due to its computational efficiency and reliable detection of community structures in large networks.

To assess the structural position of each country within the network, five centrality measures were calculated: degree centrality (number of direct connections), weighted degree centrality (strength of connections weighted by collaboration frequency), betweenness centrality (bridging role in informa-tion flow across the network), closeness centrality (average proximity to all other nodes), and eigenvector centrality (degree of connection to other influential nodes).

Topic modeling was conducted using BERTopic, which embeds documents via a pretrained BERT-based language model, reduces dimensionality through UMAP, and identifies topics through density-based clustering using HDBSCAN. Topic keywords are extracted using c-TF-IDF, reflecting how distinctively a term is used within a topic relative to others. Compared to LDA, BERTopic more pre-cisely captures semantic similarity among documents and does not require the number of topics to be specified in advance, making it well suited for exploratory research. Separate models were fitted for the two sub-periods, and each topic was labeled by the researchers based on top keywords and representative documents, with consistency between the two serving as the primary criterion for topic naming.

IV. RESULTS.

A. BIBLIOMETRIC ANALYSIS RESULTS

1) ANNUAL PUBLICATION AND CITATION TRENDS.

The 605 articles included in the analysis accumulated a total of 23,740 citations, yielding an average of 39.2 citations per paper. As shown in Fig. 1, both publication counts and citation counts follow an accelerating growth pattern in which the rate of increase itself has grown over time, with the accel-eration in citations more pronounced than that in publications. Prior to the emergence of GenAI, publications from 2016 to 2022 totaled 119 and citations totaled 2,281, whereas the cor-responding figures for 2023 to 2025 reached 486 and 21,459, representing approximately 4.1-fold and 9.4-fold increases, respectively. CPP also rose from 12.2 in 2022 to 20.2 in 2025. These findings suggest that AI trust research has deepened not only in volume but also in scholarly influence since the advent of GenAI.

2) CORE JOURNAL ANALYSIS.

The 605 articles were published across 366 journals, of which 298 (81.4%) had cited the articles. The number of publishing journals expanded approximately 3.4-fold from 91 in the period from 2016 to 2022 to 311 from 2023 to 2025, and the number of citing journals grew approximately 2.7-fold from 89 to 242 over the same period. Applying the IQR-based robust z-score criterion and restricting selection to journals ranked within the top 20 in both publication count and citation count yielded the core journals presented in Table 1.

International Journal of Human-Computer Interaction recorded the highest rankings in publication count and qualitative indicators including h-index and m-index, while Information Fusion stood out in CPP at 447.4, far exceeding the overall average of 64.9. AI & Society and Philosophy and Technology ranked consistently high across both indicators. These core journals were distributed across fields such as human-computer interaction, AI ethics and philosophy, and technology management, suggesting that scholarly interest in AI trust has emerged across a broad range of academic disciplines.

3) CORE COUNTRY ANALYSIS.

Applying the IQR-based robust z-score criterion and restrict-ing selection to countries ranked within the top 20 in both publication count and citation count yielded the core coun-tries presented in Table 2. The United States recorded the highest publication count at 167, with a total citation count of 9,353 and an h-index of 41, indicating a leading position in both research volume and scholarly influence. The United Kingdom and Germany followed in publication count, while Spain and Korea showed relatively high CPP at 93.4 and 62.8, respectively, exceeding the overall average despite compara-tively lower publication counts.

Fig. 2 presents portfolio matrices constructed with total publication and citation counts on one axis and log-difference growth rates between the pre- and post-GenAI periods on the other, with quadrant boundaries defined by the mean values of each axis. The United States was classified in the Leading quadrant in both analyses, and the United Kingdom in the Established quadrant in both. Korea and Spain were consis-tently positioned in the Rising quadrant, while Switzerland and the Netherlands fell in the Peripheral quadrant across both analyses. Germany shifted from the Leading quadrant in the publication-based analysis to the Peripheral quad-rant in the citation-based analysis, whereas China, Italy, and Australia moved from the Established or Peripheral quadrant to the Rising quadrant.

These findings suggest that while the United States and the United Kingdom have maintained their position as leading countries in AI trust research across both periods, Korea, Spain, and China appear to be countries with increasing research growth following the emergence of GenAI.

4) INTERNATIONAL COLLABORATION ANALYSIS.

Fig. 3 presents the overall international collaboration network among countries based on co-authorship relations, visualized using Gephi. The network shows a hub-oriented structure in which a small number of countries occupy central positions while many others are located at the periphery. In particu-lar, the United Kingdom, the United States, Germany, Italy, the Netherlands, Spain, Australia, Canada, and Sweden are positioned near the center of the network and maintain col-laboration links with a wide range of countries.

Table 3 presents the centrality indicators of core coun-tries that were commonly ranked in the top 20 in both publication and citation counts based on robust z-scores of centrality indices. These results are broadly consistent with the structural positions observed in Fig. 3. The United King-dom records the highest degree centrality and weighted degree centrality, indicating that it functions as the primary hub of the network in terms of both the breadth and frequency of collaborative ties. The United States shows the highest betweenness centrality (0.170), suggesting that it plays the most prominent bridging role in connecting differ-ent country groups within the network. Among the remaining countries, differentiated characteristics are observed across indicators.

Switzerland and the Netherlands were consistently classified in the Peripheral quadrant across both analyses, and Germany shifted from the Leading quadrant in the publication-based analysis to the Peripheral quadrant in the citation-based analysis. This indicates that the growth of these countries slowed relatively within the core country group. China, Italy, and Australia, by contrast, moved from the Established or Peripheral quadrant to the Rising quadrant. These findings show that while the United States and the United Kingdom maintained leading positions across both periods, Korea, Spain, and China stand out as countries with notable research growth following the emergence of GenAI.

Fig. 4 compares the international collaboration networks before and after the emergence of GenAI. The networks were constructed using countries that appeared at least twice and were connected through at least two co-authored publica-tions, and community structures were identified using the Louvain algorithm. In the earlier period, the collabo-ration network was relatively small and centered primarily around a limited number of European and North American countries, including the United Kingdom, Germany, and the United States. In the later period, the number of participating countries increased and multiple clusters emerged, reflect-ing a more extensive network structure. Compared to the earlier period, Asian countries such as Korea, China, and Japan became more visible in the later period by forming distinct collaborative clusters within the network.

European countries maintained central positions across both periods; however, some connections between European and Asian country groups were also observed in the later period, sug-gesting a gradual geographical expansion of collaborative ties.

Taken together, these results indicate that international collaboration in AI trust research is structured around a small number of central countries, including the United Kingdom and the United States, while the growing participation of

Asian countries and the partial formation of interregional connections following the emergence of GenAI point to an increasing diversification of the collaboration structure.

B. BERTopic RESULTS

This study used all-mpnet-base-v2 as the embedding model. UMAP (ncomponents=5, mindist=0.1, metric=cosine) was applied for dimensionality reduction, and HDBSCAN was used for clustering. For term extraction, a CountVec-torizer (ngramrange=, mindf=3, maxdf=0.9) was employed, and topic representation was generated using Key-BERTInspired and MMR (diversity=0.3) in combination.

Of the 605 articles, 87 (14.38%) were classified as noise, and 13 topics were identified from the remaining 518 doc-uments. The Cv coherence score of 0.5067 exceeded the threshold of 0.4 proposed by Röder et al., and the topic diversity score of 0.7615 exceeded the recommended criterion of 0.5 established by Dieng et al.. The noise ratio fell within the range of 5 to 15% recommended by Grootendorst.

1) RESEARCH TOPIC IDENTIFIED.

Table 4 presents the label, document count, proportion, and top keywords for each of the 13 topics. Topic labels were assigned by the research team based on a joint review of c-TF-IDF keywords and representative documents.

Following discussion with the aforementioned two domain experts, the 13 topics were organized into six thematic areas based on content similarity.

The first area is AI governance and trustworthiness (Topics 0, 11; 153 documents, 29.5%). Topic 0 (136 documents, 26.3%) is the largest topic overall, with accountability, robustness, and security as its central keywords. It covers research on the principles of AI trustworthiness and their connection to regulatory frameworks such as the EU AI Act. Topic 11 (17 documents, 3.3%) addresses how transparency, accountability, and data protection shape citizen trust in AI-based public administration and government services.

The second area is trust in organizational and service contexts (Topics 1, 7; 123 documents, 23.7%). Topic 1 (97 documents, 18.7%) examines trust formed in the process of accepting AI as a collaborative agent in organizational settings, with a focus on employee intentions to adopt AI and human-AI collaboration within teams. Topic 7 (26 doc-uments, 5.0%) investigates how chatbot identity design and the quality of conversational interaction influence consumer trust.

The third area is XAI and trust (Topics 2, 6; 78 documents, 15.1%). Topic 2 (51 documents, 9.8%) focuses on the evalu-ation of XAI methods such as LIME and SHAP in terms of interpretability and their effects on user trust. Topic 6 (27 doc-uments, 5.2%) examines the behavioral effects of explanation type, visualization format, and bias exposure on decision-making trust, primarily through experimental designs.

The fourth area is medical AI and trust (Topics 3, 4, 5, 9; 113 documents, 21.8%). Topic 3 (33 documents, 6.4%) addresses ethical frameworks and regulatory governance for medical AI. Topic 4 (31 documents, 6.0%) covers the techni-cal implementation of XAI in deep learning-based diagnostic models. Topic 5 (30 documents, 5.8%) examines patient and clinician intentions to accept AI doctors and the factors that shape trust in them. Topic 9 (19 documents, 3.7%) investi-gates how the design of explanations provided by AI systems influences physician trust in clinical diagnosis.

The fifth area is AI trust in education (Topics 8, 12; 33 doc-uments, 6.4%). Topic 8 (22 documents, 4.2%) addresses student and faculty trust in large language model-based tools such as ChatGPT in higher education settings. Topic 12 (11 documents, 2.1%) examines how AI literacy levels and con-cerns about AI dependency shape teacher trust in AI-EdTech in primary and secondary schools.

The sixth area is AI trust in media (Topic 10; 18 documents, 3.5%). Topic 10 covers audience trust in AI-generated news and text content and the associated risks of misinformation. Given its limited content overlap with the other areas, it was classified as a distinct thematic area.

2) CHANGES IN TOPIC DISTRIBUTION BEFORE AND AFTER.

THE EMERGENCE OF GENAI

Fig. 5 presents a portfolio matrix in which each topic is plot-ted along two axes: total document count and log-difference growth rate between the pre- and post-GenAI periods. Using the mean values of each axis (39.8 documents and 1.500 for growth rate) as boundaries, topics were classified into four types: Leading (high volume, high growth), Established (high volume, low growth), Rising (low volume, high growth), and Peripheral (low volume, low growth).

In the pre-GenAI period (2016–2022, 133 documents), Topic 0 (37.6%), Topic 1 (26.3%), and Topic 2 (15.0%) together accounted for approximately 79% of all documents. Research on AI trustworthiness principles and governance, human-AI collaboration, and XAI techniques appears to have been concentrated in a small number of areas during this period. By contrast, education-related topics (Topics 8 and 12: 3.1% combined), media (Topic 10: 3.0%), and public sec-tor topics (Topic 11: 3.0%) remained very small in number.

In the post-GenAI period (2023–2025, 385 documents), the total document count increased approximately 2.9-fold, and the relative distribution across topics also changed. The share of Topic 0 declined from 37.6% to 20.7%, and the shares of Topics 1 and 2 also decreased; however, both topics maintained above-average growth rates (1.481 and 1.504, respectively) and were classified as Leading. Topic 0, with a growth rate of 1.198 below the mean, was classified as Established.

The most notable changes were observed among the Rising topics. Topics 8 and 12 increased from a combined share of 3.1% to 7.0% and recorded the highest growth rate across all topics (2.398). Topics 10 (growth rate 1.792), 11 (1.733), and 7 (1.609) also showed above-average growth rates. The medical AI-related topics and Topic 6 were all classified as Peripheral, with below-average growth rates despite modest increases in absolute document counts. This suggests that these areas expanded at a relatively slower pace compared to other topics in the post-GenAI period, rather than reflecting an absolute decline in research interest. Topic 6 recorded the lowest growth rate of all topics (0.799). It should be noted that Topics 8 and 12 had relatively small absolute document counts, which warrants caution in inter-preting the stability of their growth rate figures.

Taken together, these findings suggest that the period fol-lowing the emergence of GenAI was associated not only with an overall increase in the volume of AI trust research but also with a shift in its thematic distribution. The relative weight of research on governance and ethics frameworks and on the behavioral effects of XAI explanations declined, while trust issues in education, media, and the public sec-tor emerged as growing areas of scholarly concern. This may reflect a gradual broadening of research interest from technical and ethical dimensions toward the social contexts in which AI is increasingly used. At the same time, XAI techniques and clinical healthcare AI topics maintained sta-ble proportions across both periods, suggesting that these areas continue to serve as enduring themes in AI trust research.

V. CONCLUSION.

A. DISCUSSION

This study analyzed 605 peer-reviewed articles on AI trust published between 2016 and 2025, combining bibliometric analysis with BERTopic-based topic modeling to compare the intellectual structure and thematic evolution of the field before and after the emergence of GenAI. The main implica-tions drawn from the findings are as follows.

First, AI trust research appears to have undergone simul-taneous expansion in volume and deepening of scholarly influence since the emergence of GenAI. Both publication counts and citation counts increased substantially in the post-GenAI period relative to the pre-GenAI period, and the rise in CPP points to the likelihood that studies produced during this period are being referenced more actively in subsequent research. These results suggest that the spread of GenAI may have influenced not only the quantitative growth of AI trust research but also the depth of scholarly engagement with the field.

Second, the core journal analysis indicates that AI trust research is distributed across multiple disciplines rather than being concentrated within a single field. Core journals were found across human-computer interaction, AI ethics and phi-losophy, and technology management, which points to the interdisciplinary character of AI trust research extending well beyond engineering-oriented system design into questions of human behavior and ethical judgment.

Third, the country-level analysis reveals that the United States and the United Kingdom maintained leading positions in both publication volume and citation impact across both periods. The United States ranked at the top in publication count and h-index and was classified as Leading in the port-folio analysis. Korea, Spain, and China, by contrast, recorded comparatively high growth rates in the post-GenAI period and were classified as Rising, reflecting a growing research interest in AI trust among these countries.

Fourth, the international collaboration network analysis confirms a hub-oriented structure centered on a small num-ber of countries. The United Kingdom recorded the highest degree centrality and weighted degree centrality, indicating its role as the primary hub in terms of both the breadth and frequency of collaborative ties. The United States showed the highest betweenness centrality, pointing to its prominent bridging function across different country groups within the network. In the pre-GenAI period, the network was rela-tively small and centered on European and North American countries; in the post-GenAI period, the number of partic-ipating countries increased and Asian countries including Korea, China, and Japan formed distinct collaborative clus-ters, reflecting a geographical broadening of the collaboration structure.

Fifth, the BERTopic analysis indicates that the thematic distribution of AI trust research shifted away from tech-nical and ethical dimensions toward the social contexts in which AI is used. In the pre-GenAI period, topics related to AI trustworthiness principles and governance, human-AI collaboration in organizations, and XAI techniques accounted for approximately 79% of all documents, suggesting a strong concentration of research in a limited number of areas. In the post-GenAI period, overall document counts increased approximately 2.9-fold, with education-related topics record-ing the highest growth rate among all topics identified, and media and public sector topics also showing above-average growth. By contrast, medical AI topics and the topic concern-ing the behavioral effects of XAI explanations were classified as Peripheral.

Although absolute document counts increased modestly, their growth rates fell below the overall average, suggesting that these areas expanded at a relatively slower pace compared to other emerging topics in the post-GenAI period.

Taken together, the findings of this study show that AI trust research has developed in the period following the emergence of GenAI alongside growth in publication volume, expansion of scholarly influence, diversification in participating coun-tries, geographical broadening of international collaboration, and shifts in thematic distribution. The observed movement of research interest from technical and ethical concerns toward social application domains such as education, media, and the public sector points to AI trust increasingly being recognized as a challenge that extends beyond system design and requires attention across a wide range of societal contexts.

B. PRACTICAL IMPLICATIONS

The findings of this study offer practical implications for aca-demic researchers, policymakers, and the field of technology development.

For academic researchers, AI trust research has expanded beyond technical and ethical dimensions since the emergence of GenAI, with a growing body of work addressing diverse application areas such as education, media, and the public sector. This suggests that AI trust needs to be examined not only as a matter of system design but also across the broader social contexts in which AI is actually used, and that more active interdisciplinary collaboration with adjacent fields such as human-computer interaction, AI ethics and phi-losophy, and technology management is needed. In particular, interdisciplinary approaches are likely to become increas-ingly important for examining the social and psychological dimensions of trust formation that are difficult to capture from within a single disciplinary perspective.

Given the pace of AI development, proactive research addressing newly emerging technical risks and trust issues is also required, along with research designs that continuously track how trust-related issues evolve alongside technological change. The interna-tional collaboration network analysis in this study found that Asian countries have formed distinct collaborative clusters following the emergence of GenAI, which points to the need to expand beyond the existing collaboration structure centered on Europe and North America toward broader inter-regional research networks. Such expansion is expected to facilitate comparative research across regions with different cultural and institutional contexts.

For policymakers, the fact that AI trustworthiness princi-ples and governance topics account for the largest share of the research reflects sustained demand for regulatory frame-works such as the EU AI Act, and continuous refinement and updating of related policies is needed. At the same time, given the rapid growth of AI trust issues in education, media, and the public sector, there is a need to move beyond broad governance discussions and develop policies and guidelines that reflect the specific characteristics of each of these areas. For instance, in education, concerns about AI literacy and AI dependency may affect trust formation, which calls for institutional support measures that address these issues. In the media sector, policy mechanisms are needed to address the risk of misinformation associated with the spread of AI-generated content.

More broadly, AI trust is an issue that requires a response beyond the level of individual countries or regions, and strengthening international governance dis-cussions and policy cooperation is required. Building flexible governance frameworks that reflect the varying cultural and normative contexts across regions is also an important task in this process.

For the field of technology development, as new technical risks such as hallucination, biased outputs, and privacy vio-lations have become prominent concerns in AI trust research following the spread of GenAI, addressing these risks sys-tematically at the stages of system development and service provision is required. The fact that AI trust research is being conducted across diverse application areas including chatbots and conversational AI services, medical AI, and AI-generated content indicates that trust issues may arise in different ways depending on the application area, and trust-aware design that reflects the characteristics of each area is needed in the development and service provision process.

In human-AI collaboration environments within organizations, employees’ intentions to adopt AI and the formation of trust during collaboration may be linked to the practical outcomes of system deployment, which means that system design should take the user-side trust formation process into account. As AI services increasingly extend into the everyday environments of general users, design perspectives that reflect the trust formation process of general users, not only expert groups, are also becoming more important. Furthermore, account-ability, robustness, and security of AI systems are technical requirements for ensuring trustworthiness, and it is necessary to restructure development processes so that these elements are sufficiently examined and implemented from the early stages of development.

C. LIMITATIONS AND FUTURE RESEARCH DIRECTIONS

This study analyzed the intellectual structure and thematic evolution of AI trust research at a macro level; however, several directions for future research remain to be addressed.

First, comparative research across academic disciplines is needed. This study examined AI trust research as a whole, but the extent to which the social sciences and the natural sciences and engineering approach AI trust issues through different topics and methodologies has not been sufficiently exam-ined. Research that systematically compares the differences and potential convergence between these two broad fields is expected to contribute to the deepening of interdisciplinary scholarship on AI trust.

Second, comparative research across continents or coun-tries is needed. Research interest in AI trust and the emphasis placed on particular topics may vary depending on cultural and institutional environments, yet this study did not directly examine such differences. Research that systematically com-pares AI trust research across continents or countries would contribute to a better understanding of how AI trust issues are shaped by the social and institutional conditions of different regions.

Third, in-depth literature reviews on specific sub-topics within AI trust research are required. This study focused on identifying the overall intellectual structure and thematic evolution of AI trust research at a broad level, and did not con-duct detailed analyses of individual topic areas. Systematic literature reviews targeting specific application areas such as education, media, and the public sector, or focused on particular themes such as XAI and governance, should be pursued in future research.

DECLARATION OF GENERATIVE AI AND AI-ASSISTED TECHNOLOGIES IN THE WRITING PROCESS

The authors did not use any generative AI or AI-assisted tools during the preparation of this work. All writing, analysis, and interpretation were performed solely by the authors.

CONFLICTS OF INTEREST has not been published previously and is not under consideration for publication elsewhere.