Abstract
Generative artificial intelligence (Gen AI), particularly ChatGPT, is fundamentally reshaping travel planning in tourism and hospitality. Despite rapid adoption, scholarly understanding remains fragmented and theoretically constrained. This study provides the first systematic review dedicated exclusively to Gen AI in travel planning. Using the scientific procedures and rationales for systematic literature reviews (SPAR-4-SLR) and PRISMA protocols, 39 empirical studies were critically analyzed through the (theory-context-characteristics-method) TCCM framework to synthesize theoretical foundations, methodological approaches, principal findings, and research gaps. The review confirms that Gen AI enhances personalization, multilingual interaction, and decision efficiency, while significantly influencing traveler cognition, preferences, and behavioral intentions. Trust, privacy, and perceived usefulness emerge as central mediating mechanisms within dominant frameworks such as technology acceptance model (TAM) and theory of planned behavior (TPB). However, the literature remains heavily anchored in cognitive-utilitarian paradigms, with limited integration of affective, relational, resistance-based, and socio-technical perspectives. To advance theoretical development, this study proposes an integrated conceptual framework that reconceptualizes Gen AI adoption as a multidimensional, ecosystem-embedded process. By bridging fragmented theoretical streams and identifying underexplored linkages, the review establishes a stronger platform for cumulative scholarship. The study also outlines governance and design imperatives to support calibrated trust and sustainable value creation in AI-enabled travel planning.
Introduction
The influence of generative artificial intelligence (Gen AI) in trip planning is rapidly reshaping the tourism and hospitality industry by transforming the way travelers plan, book, and experience their journeys. Crowley et al. (2024) report that the use of generative AI tools for travel planning is on the rise, with adoption having increased twofold since 2023. In the United States (US), digital travelers leverage Gen AI tools for leisure trips to discover destinations and activities, research flights, compare options, and construct itineraries (Statista, 2024).
Beyond simplifying logistics, Gen AI actively shapes traveler behavior, influencing preferences, consumption patterns, and decision-making processes across the tourism sector. Gen AI tools support travelers across all stages of their journey, from pre-trip planning and itinerary recommendations to in-trip guidance, translation, and post-trip evaluations (Christensen et al., 2025; Kim, Kim, Kim and Hailu, 2023; Kundan et al., 2024; Tosyali et al., 2025; Wong et al., 2023). ChatGPT, in particular, addresses key challenges in travel planning by providing immediate access to information, generating personalized itineraries, and delivering autonomous decision-support capabilities (Li and Lee, 2024; Wong et al., 2023). In addition, it assists travelers in making informed decisions regarding routes, transportation modes, and pricing options (Dwivedi et al., 2023), thereby, enhancing both the efficiency and overall quality of travel experiences.
In recent years, empirical research has increasingly investigated the potential of Gen AI in travel planning. For instance, Arora et al. (2025) examine the antecedents of the flow experience and its impact on users’ continuance intention to use ChatGPT for travel planning, while Kundan et al. (2024) explore how ChatGPT’s translation capabilities affect travelers’ behavioral intentions when encountering language barriers during international trips. These studies highlight the diverse ways in which Gen AI supports decision-making across different stages of the travel journey. However, the insights they provide remain fragmented and lack integration. Without a systematic synthesis, it is difficult to establish patterns, reconcile inconsistencies, or identify gaps within this emerging body of evidence. Therefore, a structured review is essential to consolidate existing knowledge, delineate the scope of current contributions, and provide a foundation for advancing both scholarly inquiry and industry practice.
Although several reviews have explored generative AI in tourism and hospitality, their scope has largely remained conceptual and generalized, often addressing the industry at large rather than the specific context of travel planning (Carvalho & Ivanov, 2023; Dogru et al., 2023; Dwivedi et al., 2023; Gursoy et al., 2023; Rather, 2025; Wong et al., 2025). Consequently, early discussions of ChatGPT’s potential remain insufficiently supported by empirical validation, particularly with respect to how travelers interact with AI tools during the planning process. This gap underscores the need for a focused systematic review that not only synthesizes emerging empirical evidence but also evaluates the extent to which early conceptual claims are substantiated in practice. By concentrating on travel planning, the present review complements prior overviews and extends the conversation from abstract theorizing to grounded, evidence-based insights with direct implications for both researchers and practitioners.
This study, therefore, aims to synthesize and critically evaluate existing knowledge on the use of Gen AI in travel planning, offering an integrated review that can guide both scholars and practitioners in understanding the opportunities and challenges posed by this emerging technology. Specifically, the paper seeks to answer four research questions: 1. What are the prevailing publication trends, leading journals, prominent scholars, and most influential contributions in the literature on Gen AI applications in travel planning? 2. What theoretical frameworks have been employed to study the integration of Gen AI in travel planning? 3. What antecedents, mediators, moderators, and outcomes have been identified and conceptualized in the existing body of literature on Gen AI in travel planning? 4. What are the current research gaps and proposed future directions of Gen AI in travel planning?
Methodological fragmentation continues to limit the reliability of existing reviews on generative AI in tourism and hospitality. Prior studies vary widely in scope and rigor, with some relying on bibliometric mapping (H. Li et al., 2025a; Tuyen et al., 2025), others offering narrative syntheses (Carvalho & Ivanov, 2023; Gursoy et al., 2023; Rather, 2025), and only a handful employing systematic literature review protocols with explicit Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) adherence (Fouad et al., 2024; Saleh, 2025). To address these shortcomings, the present study adopts a multi-framework approach that strengthens both rigor and transparency. Specifically, the Scientific Procedures and Rationales for Systematic Literature Reviews (SPAR-4-SLR) protocol (Paul et al., 2021) structures the review process from identification to synthesis; the theory-context-characteristics-method (TCCM) framework (Paul and Rosado-Serrano, 2019), provides a systematic lens for evaluating theoretical foundations, contextual settings, methodological approaches, and empirical characteristics; and the PRISMA guidelines (Moher et al., 2009) ensure transparency, replicability, and reporting consistency. By combining these complementary frameworks, this review not only raises evidentiary standards but also delivers a more robust, comprehensive, and practically relevant synthesis of how ChatGPT is reshaping the consumer-facing process of travel planning.
This study makes several contributions. First, the study consolidates fragmented and, at times, contradictory findings to provide a comprehensive and holistic overview of Gen AI adoption in travel planning, encompassing both its opportunities and challenges. By employing a structured review approach, it offers deeper theoretical and methodological insights, while the descriptive analysis identifies emerging patterns that capture the current state of knowledge in the field. Second, this study develops an integrated conceptual framework that delineates key antecedents, mediators, moderators, and outcomes relevant to the adoption and use of Gen AI in travel planning. Finally, by systematically identifying research gaps and proposing a future agenda, it offers a clear roadmap for advancing scholarly inquiry and guiding practical applications across the tourism sector.
Conceptual background
ChatGPT as a transformative tool for travel planning
A major advancement in artificial intelligence is the emergence of Gen AI, which extends beyond earlier rule-based and predictive systems by generating novel and contextually relevant outputs, including text, images, and customized itineraries. Within the tourism and hospitality sector, Gen AI offers substantial opportunities to enhance travel planning, enrich content creation, and transform customer service. It is rapidly transforming how travelers seek, process, and act upon information, thereby, reshaping engagement, personalization, and overall experience design (Abou-Shouk et al., 2025; Wong et al., 2023).
Traditionally, travel planning has been characterized as a complex, multi-stage decision-making process that includes searching for information, evaluating alternatives, and developing itineraries. Over the past two decades, digital platforms — from online travel agencies to algorithm-driven recommender systems — have increasingly mediated these processes. Within this evolving context, Gen AI allows travelers to co-create itineraries through interactive and conversational interfaces, blending human agency with machine intelligence. This convergence enables highly personalized and experiential planning while also raising concerns related to algorithmic bias, over-reliance on data, and issues of trust and transparency (Abou-Shouk et al., 2024; Shin et al., 2023).
ChatGPT, a state-of-the-art large language model developed by OpenAI, is a prominent example of Gen AI tool applied in travel contexts (Stergiou and Nella, 2024). Its natural language processing capabilities allow it to deliver conversational, context-aware, and dynamic assistance to travelers. Travelers generally perceive ChatGPT as user-friendly and efficient, especially in streamlining decision-making, real-time language translation, around-the-clock support, and offering tailored suggestions (Batouei et al., 2025), all of which contribute to enhanced satisfaction and more informed decision-making (Shin et al., 2023). However, challenges persist, including the risk of inaccuracies in real-time data, ethical issues related to bias and transparency, and data security concerns associated with sensitive information (Abou-Shouk et al., 2024).
ChatGPT functionalities span across the pre-trip, en-route, and post-trip phases of the tourism experience (Gursoy et al., 2023; Wong et al., 2023). It supports pre-planning activities such as destination comparison, activity planning, and transportation choices, such as flights or local transfers, simplifying the decision-making process through personalized recommendations aligned with traveler preferences (Abou-Shouk et al., 2025; Shin et al., 2023; Wong et al., 2023). ChatGPT also facilitates information gathering by providing detailed cultural insights, attraction highlights, and logistical guidance through a conversational interface, making the process more intuitive and efficient (Wong et al., 2023). Likewise, it offers real-time support such as language translation, local recommendations, and contextual knowledge, for example — flora, fauna, or dining options, thereby facilitating adaptive travel planning in response to changing circumstances. When integrated with traditional search engines, these tools can enhance the quality and comprehensiveness of travel plans; however, empirical evidence suggests that their effects on efficiency and overall satisfaction may be comparatively limited (Batouei et al., 2025). During trips, ChatGPT further support travelers by assisting in the documentation and dissemination of experiences — for instance, by drafting reviews, generating social media content, or producing reflective travel narratives.
Review of systematic review studies
Comparison of existing systematic review with the present review.
Note. This table synthesizes prior systematic reviews of Gen AI in travel planning, highlighting their aims, methodologies, scopes, and contributions. The present review distinguishes itself by focusing on travel planning, consolidating empirical evidence and theories.
The body of literature reviewed underscores the growing interest in the application of Gen AI, especially ChatGPT, within the tourism and hospitality sectors. While several studies have examined the general benefits of ChatGPT, including its potential to enhance personalization, service delivery, and operational efficiency (Gursoy et al., 2023; Tuyen et al., 2025), most of these reviews remain generalized, offering high-level overviews rather than a focused exploration of its use in consumer-facing travel planning. Many studies focus on decision-making (Wong et al., 2023) or service interactions (S. Li et al., 2025b), but they fail to explore how ChatGPT influences the actual travel planning process. This suggests that while there is increasing awareness of ChatGPT’s role in the tourism sector, there remains a lack of focused studies on how it can specifically transform travel planning.
Bibliometric analyses are useful for identifying macro trends in AI adoption across various sectors (Saleh, 2025; Tuyen et al., 2025) but, but these studies fall short of offering a deep dive into the practical implications and user experiences of ChatGPT within the travel planning context. They tend to highlight broad research themes without integrating empirical data related to travelers’ behavior and decision-making, leaving significant gaps in understanding how travelers engage with Gen AI tools and what factors drive their adoption of ChatGPT for planning trips.
Unlike prior reviews that examine ChatGPT and Gen AI across tourism and hospitality broadly, for example (Tuyen et al., 2025)0 and Saleh (2025), this study isolates travel planning as a theoretically bounded, multi-stage cognitive and behavioral process encompassing information search, evaluation of alternatives, itinerary construction, risk assessment, and booking decisions. Rather than treating planning as one application among many, we position Gen AI as an active decision-shaping agent within the traveler’s pre-trip cognition. By synthesizing adoption, resistance, affective, relational, and socio-technical perspectives into an integrated framework, this study reconceptualizes AI-assisted travel planning as a multidimensional phenomenon embedded in broader service ecosystems. Furthermore, technologies such as virtual tourism and AI-mediated simulations are examined not as peripheral innovations but as mechanisms of cognitive pre-experience that influence destination image, perceived behavioral control, and intention formation. Through this process-level and theory-integrative lens, the study advances beyond descriptive sectoral mapping to provide a structured foundation for cumulative theory development in AI-enabled travel planning.
Methodology
Review protocols
To ensure rigor, transparency, and replicability, this study integrated three complementary protocols (Figure 1). The SPAR-4-SLR (Paul et al., 2021) was employed to structure the review process across the stages of assembling, arranging, and assessing. To systematically evaluate the theoretical foundations, contextual settings, empirical characteristics, and methodological approaches, the TCCM framework (Paul and Rosado-Serrano, 2019) was applied. Finally, the PRISMA guidelines (Moher et al., 2009) were adopted to guide the identification, screening, eligibility, and inclusion process, thereby enhancing the transparency and replicability of the review. Methodological integration of SPAR-4-SLR, TCCM, and PRISMA in the present review.
Assembling and identification
Two leading bibliographic databases, Scopus and Web of Science, were selected as the primary data sources. These databases were chosen because of their comprehensive coverage of peer-reviewed journals and their widespread use in bibliometric and systematic review studies in management, tourism, and information systems research (Pranckutė, 2021).
A systematic search was conducted in August 2025, limiting the search to research before this period. Moreover, the study uses a combination of Boolean operators to simplify the search process (Randles and Finnegan, 2023). The constructed search query combined terms related to “generative AI”, “travel”, and “planning”. The search string was as follows: (“generative AI” OR “generative search” OR “ChatGPT” OR “Gemini” OR “Claude” OR “Ilama” OR “mistral” OR “perplexity AI”) AND (“holiday” OR “travel” OR “tour” OR “itinerary” OR “journey” OR “pre-trip” OR “post-trip” OR “trip”) AND (“preparation” OR “planning” OR “decision-making”). The search was applied to titles, abstracts, and keywords across both databases to maximize the retrieval of relevant publications. This process yielded 179 records in and 109 records in WoS.
Arranging and screening for eligibility
To ensure quality and focus, strict eligibility criteria were applied during the screening process. Only peer-reviewed journal articles were included, while conference proceedings, book chapters, editorials, books, notes, reviews, and other non-scholarly sources were excluded. After this step, 84 articles in Scopus and 68 in WoS remained.
According to Dobrescu et al. (2021), restricting synthesis to English-language publications, often due to time and resource constraints, has minimal impact on the validity of systematic reviews. Based on this rationale, the sample was further refined to include only English-language publications, resulting in 81 records from Scopus and 56 from WoS. The two datasets were then merged in R, and after removing 53 duplicates, a final dataset of 94 unique articles was obtained.
Subsequently, each article was assessed against a set of predefined eligibility criteria to ensure both rigor and relevance. The first criterion was topic relevance, which required an explicit focus on ChatGPT or Gen AI within the travel planning context. The second was methodological scope, allowing for the inclusion of both empirical studies — quantitative, qualitative, or mixed methods — and conceptual or theoretical contributions that specifically addressed the role of Gen AI in travel planning. The third criterion was outcomes of interest, which narrowed the scope to studies offering insights into adoption, intention, or decision-making processes shaped by Gen AI tools during travel planning. In total, 30 studies satisfied all criteria.
To strengthen coverage and reduce the risk of omitting relevant studies, forward and backward referencing was also employed (Gusenbauer, 2024; Horsley et al., 2011). By tracing citations of the initially identified papers using Google Scholar and reviewing their reference lists, additional works that met the inclusion criteria were captured. This iterative approach complemented the database search strategy, ensuring a more comprehensive and accurate representation of the emerging literature. Applying these combined procedures ultimately yielded a final set of 39 studies that formed the basis of the systematic review.
Assessing and reporting
For the bibliometric analysis, the final set of studies was processed using Biblioshiny, a web-based interface of the R-package bibliometrix that facilitates comprehensive bibliometric analysis and visualization (Aria and Cuccurullo, 2017). Performance metrics were reported to capture the intellectual structure of the field, including the annual growth rate of publications, the most relevant articles, the most prolific authors, and the most influential journals contributing to the knowledge base on Gen AI in travel planning.
For the qualitative analysis, key information was systematically extracted from each included study. This comprised the theoretical framework employed (if any), the research context, for example — country, population group, the characteristics examined — including independent, dependent, mediating, moderating variables, the methodological approach adopted — such as qualitative, quantitative, conceptual, mixed methods, and the key findings together with their practical and theoretical implications.
Following this extraction, a narrative synthesis was undertaken to identify converging themes, research gaps, and potential directions for future inquiry. To strengthen the analysis, the insights were further mapped onto the TCCM framework, enabling a systematic evaluation of the state of knowledge and guiding the development of a future research direction.
Findings, discussion, and future research
Review protocols
The bibliometric dataset reflects a rapidly emerging research stream on Gen AI in tourism and hospitality. All 39 publications fall within the timespan 2023 to 2025, underscoring the field’s novelty and explosive growth, with an annual growth rate of 274.17%. This suggests that Gen AI research in tourism and hospitality has largely been driven by the rapid spread of technologies such as ChatGPT, with theoretical development to date being more reactive than foundational. Consequently, there is a clear need for theory extension and theory building that explicitly accounts for AI agency, human–AI co-decision-making, and algorithmic uncertainty within tourism contexts.
The average document age of just 0.366 years confirms the recency of contributions, yet the citation rate (16 per article) indicates strong scholarly attention and early influence. This suggests strong industry and policy relevance. However, the speed of publication also raises concerns about premature managerial prescriptions, reinforcing the need for more longitudinal, field-based, and experimental studies to support evidence-based adoption. Thus, practitioners are advised to adopt generative AI cautiously (Dwivedi et al., 2025), addressing transparent use and human oversight, particularly in areas where trust, accuracy, and ethical responsibility directly influence tourist decision-making.
Most relevant journals
Leading journals contributing to Gen AI research in travel planning.
Research has been published across 15 different journals, suggesting that studies are diffusing across multiple disciplinary domains. The dataset confirms that research on Gen AI for travel planning is at an infant but booming stage — global, collaborative, and already impactful. The distribution of articles reveals that research on generative AI in travel planning is concentrated in higher-ranked outlets, reflecting the field’s growing academic recognition, yet it also shows a degree of dispersion across lower-tier journals, underscoring its still-emerging and exploratory nature.
The publication landscape reveals that research on Gen AI in travel planning is mostly concentrated in Q1 outlets. TR leads with six articles, followed closely by CIT, IJHM, and JTR, each publishing four. The prominence of A* and A-ranked journals such as IJHM and JTR underscores the academic legitimacy and growing importance of this field within tourism and hospitality research. While the bulk of publications are anchored in core tourism and hospitality journals, a smaller but noteworthy set of studies appears in interdisciplinary outlets such as IJHCI, IR, and Sustainability. This suggests opportunities for cross-pollination of theories from information systems (IS), human-computer interactions (HCI), and management studies into tourism research, developing theories that account for tourism-specific characteristics such as experiential consumption, uncertainty, emotional decision-making, and ethical responsibility toward destinations and communities. From a practical standpoint, the presence of Gen AI research in sustainability- and technology-oriented journals signals that industry stakeholders are concerned not only with efficiency and personalization, but also with governance, transparency, and societal impact. This reinforces the need for tourism managers and policymakers to view Gen AI adoption as a strategic capability requiring ethical oversight (Kim, Kim, Kim and Kim, 2023).
Most influential authors
Figure 2 presents the key authors contributing to the emerging body of research on generative AI in travel planning. This depicts that a small but active group of scholars. These scholars have played a pivotal role in advancing both conceptual and empirical knowledge, shaping how the field is developing. Key authors contributing to research in Gen AI for travel planning.
Kim J leads the field with seven publications, followed by Kim MJ with five, both contributing extensively to research on technology adoption and user experience. Han H, Kim JH, Kim S, and Seyfi S have each authored four papers, underscoring the prominence of South Korean scholars while simultaneously highlighting Seyfi’s international contributions. Hailu TB and Koo C, with three articles each, extend the discourse by linking travel planning to digital innovation and information systems, whereas Abou-Shouk M and Al-Ansi A, with two publications apiece, introduce perspectives from Middle Eastern and global tourism contexts.
From a theoretical perspective, the field risks relying on the same ideas, methods, and assumptions. While this has helped early development, it highlights the need for more diverse perspectives. From a practical perspective, the growing geographic diversification of contributors is particularly significant, suggesting increasing recognition that Gen AI adoption is different contexts. Future collaborative research can therefore play a key role in informing regionally appropriate Gen AI strategies rather than universal best practices.
Overall, these authors not only account for a significant share of publication volume but also provide essential theoretical and empirical foundations for advancing understanding of Gen AI in travel planning.
Most relevant articles
Most influential papers advancing to Gen AI research in travel planning.
The most cited works emphasize trust and ethical concerns (Ali et al., 2023; Kim, Kim, Park, et al., 2023), accuracy and hallucinations (Christensen et al., 2025; Kim, Kim, Kim and Kim, 2023), and decision-making in travel planning (Kim, Kim, Kim and Hailu, 2023; Shin et al., 2023). Other contributions expand the scope through technology acceptance (Solomovich and Abraham, 2024), parasocial interaction (Duong et al., 2024), and smart tourism ecosystems (Suanpang and Pothipassa, 2024). Overall, these studies highlight a dual narrative: ChatGPT enhances personalization and engagement, but issues of trust, misinformation, and ethical responsibility remain central research challenges.
Key thematic clusters of generative AI and travel planning research
To better understand how research on generative AI in travel planning is structured, a co-occurrence network analysis was conducted. This approach identifies the most frequently associated keywords across the literature and organizes them into thematic clusters, highlighting both conceptual linkages and research priorities. The co-occurrence network (Figure 3) reveals three main thematic clusters underpinning research on generative AI in travel planning. Co-occurrence network of generative AI in travel planning research.
Cluster 1: User trust and perceived value
This cluster centers on travel planning and its associated constructs of trust, privacy, and usefulness. Among these, trust emerges as the most influential, suggesting it acts as a key bridge between travel planning and broader AI adoption discussions. Prior work shows that trust is critical in mediating how users perceive system attributes such as credibility, intelligence, and relevance, ultimately shaping their willingness to rely on AI tools in tourism contexts (Ali et al., 2023). Privacy concerns, while often seen as barriers, indirectly influence behavioral intentions by shaping attitudes toward AI-powered travel tools (Chauhan and Jishtu, 2025). Moreover, algorithm aversion studies suggest that when ethical or quality issues arise, users’ trust can diminish significantly, limiting acceptance of ChatGPT-generated travel recommendations (Kim, Kim, Kim and Kim, 2023). Collectively, these findings highlight that while perceived usefulness motivates adoption, trust functions as the essential mechanism that translates perceived value into actual behavioral outcomes.
Cluster 2: ChatGPT as the core technological hub
This cluster is dominated by ChatGPT and generative AI, positioning them as central nodes that connect diverse concepts such as artificial intelligence, chatbots, technology acceptance model (TAM), continuance usage intention, and tourist decision-making. ChatGPT’s centrality reflects its ability to integrate personalization, efficiency, and scalability across multiple tourism domains, from customer engagement to service quality (Li and Lee, 2024; Wong et al., 2025). Adoption studies grounded in TAM and unified theory of acceptance and use of technology (UTAUT) emphasize perceived usefulness, ease of use, and enabling conditions drive tourist acceptance of AI chatbots, with ChatGPT emerging as a particularly strong predictor of continuance usage (Duong et al., 2024; Han et al., 2024). Additionally, relational perspectives such as parasocial interaction theory (Duong et al., 2024) and cognitive-affective-normative models (Han et al., 2024) underline how ChatGPT fosters ongoing engagement through emotional and normative pathways. These converging strands establish ChatGPT not merely as a functional tool but as the central anchor in AI discourse, shaping both adoption and resistance patterns in travel planning.
Cluster 3: Behavioral theories of Gen AI adoption in travel planning research
This cluster is smaller but introduces conceptual depth, with nodes like innovation resistance theory and technology adoption offering alternative frameworks to examine user acceptance and resistance behaviors. Extended behavioral models, such as the theory of planned behavior (TPB), have incorporated AI-specific constructs like anthropomorphism and privacy to explain user attitudes and adoption intentions in travel planning (Chauhan and Jishtu, 2025). Expectancy–disconfirmation theory further illustrates how users’ satisfaction with AI recommendations depends on perceived effort and response quality (Kim et al., 2024). Innovation resistance theory, though less commonly applied, provides valuable insights into how perceived risk, uncertainty, and skepticism limit adoption, particularly among more cautious user segments (Siamak Seyfi, Changkyu Lee et al., 2025). Complementary perspectives such as parasocial interaction (Duong et al., 2024) and pleasure-arousal-dominance (PAD) theory (Xu et al., 2025) demonstrate that affective and relational responses to generative AI tools are equally significant alongside cognitive evaluations. Overall, this cluster signals an evolution from simple adoption models to more nuanced behavioral theories that capture both acceptance and resistance in generative AI travel research.
TCCM analysis and outcomes
Insights from TCCM evaluation.
Theoretical insights and future direction
Research on generative AI in tourism is heavily anchored in established frameworks such as the TAM (Abou-Shouk et al., 2025; Christensen et al., 2025; S. Li et al., 2025b; Li and Lee, 2024), the TPB (Chauhan and Jishtu, 2025; Han et al., 2025) and their extensions. These models consistently highlight constructs like trust, usefulness, ease of use, enjoyment, innovativeness, and privacy concerns as drivers of adoption (Ali et al., 2023; Duong et al., 2024). More nuanced theoretical applications include parasocial interaction theory (Duong et al., 2024), affordance-actualization theory (Li and Lee, 2024), innovation resistance theory (Siamak Seyfi, Abolfazl Siyamiyan Gorji et al., 2025; Siamak Seyfi, Changkyu Lee et al., 2025), and experience-based frameworks such as PAD theory (Xu et al., 2025).
However, theoretical diversity remains limited, with most studies reinforcing existing adoption and continuance pathways. Future research should integrate ethics-driven perspectives (e.g., fairness, bias, sustainability), human–AI collaboration theories, and sociocultural frameworks to capture the broader implications of AI adoption. For example, incorporating institutional theory could explain how regulations shape adoption, while trust repair theories may clarify user responses to AI errors (Kim, Kim, Kim and Kim, 2023).
Moreover, cross-theoretical integrations, such as combining TPB with parasocial interaction or innovation resistance, would enrich understanding of both acceptance and resistance behaviors in AI-mediated travel planning. While TAM and TPB dominate the literature on Gen AI adoption in tourism and travel planning, their main focus on cognitive assessment limits their ability to capture affective, relational, and socio-technical dynamics that are increasingly evident in recent studies. Instead of viewing alternative theories as replacements, this review combines them as complementary perspectives. Specifically, TAM and TPB address instrumental and normative evaluations; parasocial interaction theory and PAD explore relational and emotional engagement with conversational agents; innovation resistance theory explains rejection and skepticism; and affordance-actualization theory highlights how system capabilities and design features influence user perceptions. By integrating these theoretical frameworks, the review presents a coherent, multi-level understanding of Gen AI adoption and resistance within travel planning contexts.
Contextual insights and future direction
Empirical studies have been geographically diverse, with strong regional representation from Asia including China, South Korea, India, Vietnam, Malaysia, Japan, and Turkey (Bui et al., 2025; Fakfare et al., 2025, 2025b; Han et al., 2025; Seyfi, Kim, et al., 2025; Seyfi, Lee, et al., 2025), North America (US, Canada) (Liu and Shrum, 2002; Siamak Seyfi, Changkyu Lee et al., 2025; Topsakal, 2025), and parts of Europe and the Middle East (UK, Israel, UAE, Oman, Egypt) (Siamak Seyfi, Abolfazl Siyamiyan Gorji et al., 2025). Contextually, research spans tourism planning, hospitality, translation services, virtual tourism for older adults, and travel confidence building (Kundan et al., 2024; Liu et al., 2025). Yet, most studies are concentrated in urban, digitally advanced, or tourism-heavy regions, with limited focus on underrepresented geographies such as Africa, Latin America, and the Global South region.
Future work should prioritize cross-national comparative studies to test the cultural robustness of adoption models, and extend to vulnerable groups like elderly travelers, disabled tourists, or low-digital-literacy users (Suanpang and Pothipassa, 2024). Furthermore, sustainability-driven contexts, such as responsible travel, ecotourism, and climate-conscious trip planning, remain underexplored. Embedding AI adoption within regional cultural norms, digital divides, and regulatory landscapes would provide richer contextual insights and broaden the global generalizability of findings.
Characteristics insights and future direction
Across the studies, antecedents such as trust, usefulness, enjoyment, innovativeness, anthropomorphism, and privacy have dominated the research agenda (Ali et al., 2023; Chauhan and Jishtu, 2025). Mediators include trust (Li and Lee, 2024; Siamak Seyfi, Changkyu Lee et al., 2025), attitudes (S. Li et al., 2025b), satisfaction (Duong et al., 2024), and inspiration (Wang et al., 2025). Moderators such as age, generation, personality traits, and cultural differences have also been tested (Topsakal, 2025). The dependent variables largely focus on adoption, continuance intention, word-of-mouth, loyalty, and resistance (Han et al., 2024; Xu et al., 2025).
Despite this progress, most studies remain technology-centric, emphasizing adoption outcomes at the expense of well-being, empowerment, and socio-ethical outcomes. For instance, while Liu et al. (2025) linked Gen AI avatars to older adults’ well-being, broader connections between AI adoption and traveler equity, accessibility, sustainability, or psychological resilience remain scarce. Future research should expand outcome variables to include destination choice, trust repair, digital confidence, emotional well-being, and sustainability-driven behaviors. Examining value co-creation and destruction (Bui et al., 2025) could also reveal how generative AI reshapes the tourist experience beyond simple adoption.
Methodological insights and future directions
The methodological profile of this field is dominated by quantitative surveys analyzed via SEM/PLS-SEM (Abou-Shouk et al., 2024; Han et al., 2025), supported by fsQCA and ANN-based approaches (Foroughi et al., 2025). Controlled experiments have been employed to examine message framing, trust, and response time (J. Kim et al., 2025a; Kim et al., 2024; Kim, Kim, Kim and Kim, 2023; Kim, Kim, Kim and Hailu, 2023; Kim, Kim, Park, et al., 2023; Kumamoto and Joho, 2025; Shin et al., 2023; Wong et al., 2025). Some studies adopt qualitative interviews or typological analyses (J. Kim et al., 2025a; Wong et al., 2025). Current approaches often rely on cross-sectional, self-reported data and scenario-based settings.
Future research should embrace longitudinal designs to track evolving trust and adoption over time (Chauhan and Jishtu, 2025), field experiments and real-world behavioral data (e.g., OTA log files, clickstream analysis) to capture authentic usage. Incorporating multi-method triangulation (survey + experiments + ethnography) would strengthen validity. Moreover, integrating immersive designs (VR, AR, avatar-guided travel) and AI–IoT integration studies (Bui et al., 2025; Suanpang and Pothipassa, 2024) would allow researchers to test adoption in more realistic environments. Finally, expanding cross-cultural longitudinal panels can reveal how adoption pathways differ across global regions and evolve over repeated interactions.
Conclusion
Theoretical contributions
From a theoretical perspective, the findings highlight the centrality of trust, privacy, and perceived usefulness as mediating constructs within established adoption frameworks such as TAM and TPB (Chauhan and Jishtu, 2025). However, the analysis also reveals a growing need to expand beyond these dominant paradigms by incorporating complementary perspectives such as innovation resistance theory and socio-technical systems theory. This extension will allow future research to capture both drivers of adoption and barriers to resistance, thereby offering a more holistic understanding of user interactions with generative AI in tourism contexts. Moreover, the clustering results suggest that generative AI adoption is not only a technological process but also a behavioral and relational phenomenon, where anthropomorphism, perceived value, and cultural context play vital roles (Chauhan and Jishtu, 2025; Wong et al., 2025). Furthermore, this study provides a theoretical contribution by integrating cognitive, affective, relational, and socio-technical perspectives into the Gen AI adoption literature. Unlike earlier reviews that primarily list dominant frameworks such as TAM and TPB, this study shows how these models serve as a cognitive foundation, often enhanced by affective (PAD), relational (parasocial interaction), and resistance-focused views. By combining these theories into a unified framework, future research can better understand their complementary functions and identify less-explored theoretical links, providing a basis for more comprehensive theory development in Gen AI and tourism research.
Managerial implication
From a managerial standpoint, the results provide actionable insights for tourism and hospitality stakeholders. First, the emphasis on trust and privacy signals the importance of transparent data practices, ethical AI design, and user empowerment in building long-term acceptance. In practical terms, this can be achieved by clearly communicating how customer data are collected and used within AI-driven travel planning tools, embedding privacy-by-design features, and providing users with visible control over personalization settings. For example, tourism firms integrating ChatGPT-like systems into their websites, mobile apps, or OTA interfaces can include explanations of recommendation logic, disclaimers about potential limitations, which have been shown to reduce perceived risk and increase trust in AI-supported decisions (Christensen et al., 2025).
Second, as ChatGPT and similar tools become the central hub for trip planning and customer engagement, firms should leverage these technologies to deliver personalized, co-creative, and contextually adaptive experiences that strengthen customer loyalty. This can be operationalized by integrating generative AI with booking engines, CRM systems, and destination content, allowing recommendations to dynamically adjust based on traveler preferences and situational context. Hybrid AI–human planning models, where AI-generated itineraries are reviewed or refined by human experts, can further mitigate risks associated with misinformation while preserving efficiency and personalization (Kim, Kim, Kim and Hailu, 2023).
Finally, by understanding the dual forces of adoption and resistance, managers can proactively design interventions, such as trialability features, educational campaigns, and user-friendly interfaces that reduce uncertainty and encourage responsible use. Rather than relying solely on generic user education, firms can embed learning mechanisms within the customer journey, for instance, interactive prompts that guide users in refining preferences, comparison tools that display multiple itinerary options, or sandbox environments that allow experimentation with AI planning without immediate commitment. Such design-based interventions support calibrated trust and more informed decision-making, which are especially important in high-involvement tourism contexts (Ali et al., 2023).
Limitations and future research directions
Despite the contributions and implications, this study has several limitations that reflect both the novelty of the field and the structure of the existing evidence base. First, Gen AI research in tourism is still in its early stages, which constrains theoretical maturity and cumulative consolidation. Many studies remain exploratory and adoption-focused, limiting deeper theory building. Second, the review draws exclusively on peer-reviewed journal articles, which may introduce publication bias and exclude valuable insights from industry reports, conference proceedings, and working papers where technological experimentation often emerges first. Third, the geographic distribution of studies is uneven, with a concentration in Asian and digitally advanced markets, potentially limiting the cross-contextual generalizability of findings. Most critically, the empirical literature is methodologically skewed toward cross-sectional surveys and vignette-based experiments. While appropriate for early-stage inquiry, these designs restrict causal inference, heighten common method bias, and may overestimate adoption intentions while underrepresenting real-world uncertainties such as misinformation, evolving trust, and algorithmic opacity in high-stakes planning contexts.
Future research should therefore prioritize methodological diversification and rigor. Longitudinal designs tracking travelers across planning stages can illuminate how trust, reliance, and resistance evolve over time. Field-based and quasi-experimental studies embedded within real travel platforms would enhance ecological validity by capturing observable behavioral outcomes. Integrating behavioral log data with perceptual measures can reduce self-report bias, while cross-cultural comparative research is essential to account for institutional and socio-cultural variation. Advancing toward multi-method, real-world research designs will be critical for strengthening causal explanations and supporting responsible, evidence-based AI deployment in travel planning.
Footnotes
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Author biographies
Appendix A. Papers included in this study
| No | Author(s) (year) | AI tool | Theory/model | Country | Participants | Method | Key findings | Theoretical implications | Managerial implications | Limitations/future research |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Christensen et al. (2025) | ChatGPT | TAM, TPB | North America & Asia-Pacific | 900 consumers | Survey (both PLS-SEM & CB-SEM) | Prior AI use ↑ familiarity → ↑ perceived ease, usefulness, control, trust; hallucinations acknowledged but adoption persists. | Extends TAM/TPB by adding prior use, familiarity, hallucination awareness. | Build trust, transparency, and educate users about hallucinations. | Limited to TAM/TPB; future work on trust paradox, ethics, and generational cohorts. |
| 2 | Topsakal (2025) | Gen AI | Extended TAM | US | 387 participants | Survey (PLS-SEM) | Familiarity & trust strongly predict intention; ease of use less relevant; age moderates usefulness effect. | Challenges TAM by emphasizing trust over ease of use. | Educate users, emphasize benefits, tailor by age group. | US-only, self-report bias; cross-cultural and alternative models needed. |
| 3 | Kim et al. (2024) | ChatGPT | CASA paradigm | Not specified | 774 Travelers (3 studies) | Experiments (ANOVA) | Explanatory style ↑ acceptance of recommendations. | Advances chatbot theory by showing style matters. | Use explanatory messaging, avoid overload, tailor to user familiarity. | Limited to destination recommendations; more contexts and personalization needed. |
| 4 | Stergiou and Nella (2024) | ChatGPT | ADT | Not specified | 360 units of ChatGPT outputs | Qualitative (Thematic & content analysis) | AI provides context-relevant, personalized advice. | Extends ADT to AI, reframing accessibility/diagnosticity. | Design context-aware, adaptive advisory tools. | Limited by predefined queries; future work on diverse contexts and ethics. |
| 5 | Bui et al. (2025) | Gen AI (ChatGPT) | SDL, SPT | UK & Turkey | Qualitative phase (18); quantitative phase (382) | Qualitative (Thematic analysis) + survey (SEM) | ChatGPT adds efficiency (value co-creation) but inaccuracies erode trust (value co-destruction). | Reframes TAM/SDL: usefulness central, ease of use less relevant. | Improve accuracy, manage trust, refine services. | Focused on ChatGPT and limited contexts; broader Gen AI needed. |
| 6 | Siamak Seyfi, Changkyu Lee et al. (2025) | Gen AI (ChatGPT, GPT-4, Bard, Cohere, and Stable diffusion) | IRT, generation theory | South Korea & US | 628 valid respondents | Survey (PLS-SEM, MGA) + fsQCA | Barriers (usage, risk, image) ↓ trust → ↓ recommendation; generational differences evident. | Integrates IRT with generational lens. | Tailor adoption strategies by cohort, such as Gen Z needs transparency, Boomers want reliability). | Limited barriers studied; broader factors (ethics, culture) needed. |
| 7 | Han et al. (2025) | ChatGPT | TAM, TPB, BRT | South Korea | 464 valid respondents | Survey + fsQCA | Adoption driven by combinations of usefulness, control, convenience, norms; barriers only matter when enablers are weak. | Shows configurational, multi-theory drivers of adoption. | Provide personalized, credible, and private AI services. | South Korea-only, single stakeholder; future work cross-cultural, multi-stakeholder. |
| 8 | Siamak Seyfi, Abolfazl Siyamiyan Gorji et al. (2025) | Gen AI (ChatGPT) | IRT | Iran | 22 outbound travelers | Interviews (Thematic analysis) | Resistance shaped by functional & psychological barriers; typology: rejecters, postponers, opinion leaders. | Extends IRT with typology, reframes image barriers as identity-based. | Address resistance types via hybrid AI-human models, transparency, cultural sensitivity. | Small, single-country sample; resistance over time needs study. |
| 9 | Siamak Seyfi, Myung Ja Kim et al. (2025c) | Gen AI (ChatGPT, GPT-4, Bard, Cohere, and Stable diffusion) | IRT, Big Five personality traits | South Korea & US | 628 valid travelers | Survey (SEM, MGA) + fsQCA | Usage/value barriers & anxiety ↓ trust; personality traits shape adoption paths. | Integrates IRT with personality theory. | Tailor systems to personality-driven preferences, ensure transparency. | South Korea/US only; focus on outbound travelers. |
| 10 | Fakfare et al. (2025a) | ChatGPT | Innovation characteristics | South Korea | 505 valid respondents | Survery (SEM) + fsQCA | ChatGPT’s benefits, compatibility, and trialability → adoption → positive WOM. | Shifts adoption theory from linear to flexible state-based models. | Highlight benefits, simplify use, integrate with bookings. | ChatGPT-only, excludes non-users and other platforms. |
| 11 | Li and Lee (2024) | Gen AI (ChatGPT) | Affordance-actualization theory, communication theory | South Korea | 477 valid travelers | Survey (SEM) | Timeliness, personalization, and anthropomorphism drive trust; cognitive trust, not emotional trust, predicts loyalty and adoption. | Extends algorithm aversion and AAT by showing communication quality fosters trust and loyalty. | Improve timeliness, accuracy, personalization, and anthropomorphic features in ChatGPT-based travel planning. | Limited to pre-travel stage; future studies should use mixed methods and cover full travel journey. |
| 12 | Shin et al. (2023) | ChatGPT | Choice overload effect | US | 1362 Mturk respondent (5 studies) | Experiments (ANOVA) | Automated option reduction by ChatGPT reduces satisfaction, trust, and intention; human involvement mitigates negative effects. | Extends choice overload theory to AI-mediated contexts, emphasizing the role of trust and autonomy. | Use ChatGPT to generate broad option sets, encourage hybrid human–AI decision-making. | Scenario-based experiments limit generalizability; future research should test in real-world contexts. |
| 13 | Abou-Shouk et al. (2025) | ChatGPT | Extended TAM | Egypt | 514 hospitality customers | Survey (PLS-SEM) | Perceived enjoyment, ease of use, and usefulness shape attitudes, which strongly predict intention and customer value. | Extends TAM by integrating hedonic and value-creation dimensions. | Design enjoyable, easy-to-use ChatGPT systems, integrate with human service, and use feedback-driven improvements. | Limited to Egypt; intention-based measures only; cross-cultural and behavioral studies needed. |
| 14 | S. Li et al. (2025b) | ChatGPT | Extended TAM | China | 491 valid tourists | Survey (CB-SEM) | Ease of use, usefulness, and enjoyment enhance attitudes, which drive behavioral intention. | Extends TAM by integrating rational (utilitarian) and emotional (hedonic) factors. | Enhance usability, reliability, and enjoyment of ChatGPT; foster trust with transparent design. | Limited to self-reported data and a single context; future research should use behavioral data and varied tourism settings. |
| 15 | Kim, Kim, Park, et al. (2023) | ChatGPT | Accessibility-diagnosticity framework | US & UK | 1547 participants (5 studies) | Experiments (ANOVA) + logistic regression | Inaccurate ChatGPT info reduces trust, accuracy, and visit intention, especially when errors are salient/relevant. | Extends AI adoption research by identifying boundary conditions of misinformation in AI. | Minimize errors, implement verification systems, reduce salience of inaccuracies. | Scenario-based, not real-world; future research should test long-term impacts and cross-cultural effects. |
| 16 | Foroughi et al. (2025) | ChatGPT | Extended UTAUT2 | Malaysia | 410 travelers | Survey (PLS-SEM) + ANN | Usefulness, enjoyment, facilitation, and innovativeness drive adoption; risk aversion hinders it. | Extends UTAUT2 by adding innovativeness and risk aversion; shows diminished role of social influence. | Highlight usefulness and enjoyment, target innovative users, reduce risk perceptions with transparency. | Limited to Malaysia and early adoption stage; cross-cultural and psychological studies needed. |
| 17 | Fakfare et al., 2025b | Gen AI (ChatGPT) | TPB, cognitive-emotional-behavioral framework, complexity theory | South Korea | 505 valid respondents | Survey (SCA + NCA regressions) + fsQCA | Loyalty is shaped by the interplay of risks variables such as cognition, and emotional responses. | Advances loyalty theory by showing complex, asymmetric pathways (equifinality). | Reduce time/privacy risks, design empathetic AI interfaces, build emotional well-being. | South Korea only; lacks longitudinal insights; future studies should explore trust, ethics, and cross-cultural variation. |
| 18 | Ghorbanzadeh et al. (2025) | ChatGPT | UTAUT, experiential consumption theory | Not specified | 384 valid travelers | Survey (PLS-SEM) | Both utilitarian and hedonic values drive adoption; performance expectancy and social influence matter, facilitating conditions do not. | Extends UTAUT by positioning hedonic and utilitarian values as antecedents. | Create efficient, enjoyable itinerary tools; enhance usability with AR and intuitive design. | Limited by self-reported data, traveler-only perspective, and focus on ChatGPT; future research should broaden scope. |
| 19 | Batouei et al. (2025) | ChatGPT | Extended TAM | Malaysia | 441 travelers | Survey (PLS-SEM) | Attitude strongest predictor of adoption; perceived enjoyment and convenience drive positive attitudes; usefulness is a less significant driver. | Extends TAM by integrating hedonic/contextual factors; challenges traditional role of innovativeness. | Emphasize enjoyment, convenience, and info quality; market ChatGPT as engaging travel companion. | Malaysia-only sample; some unsupported hypotheses; future studies should explore ethics, transparency, and trust. |
| 20 | Kumamoto and Joho (2025) | Gen AI (ChatGPT) | Collaborative info seeking; human–AI collaboration framework | Japan | 20 students (10 pairs) | Experiment + survey (t-tests and Wilcoxon signed-rank tests) | ChatGPT improved completeness and detail but not speed/satisfaction; better for ideation than logistics. | Demonstrates AI’s supportive role in collaborative ideation but limits in logistics-heavy tasks. | Use ChatGPT to support idea generation; integrate with live data for real-time accuracy. | Small student sample, limited scope; needs diverse participants and longitudinal designs. |
| 21 | Suanpang and Pothipassa (2024) | Gen AI (ChatGPT, NLP, IoT) | Smart tourism & sustainable tourism frameworks, technology acceptance & human–AI interaction theories, service innovation models | Thailand | 416 tourists, 20 IT experts | Agile methodology, expert testing, survey (SEM) | Integrating GenAI, NLP, and IoT improved personalization, accessibility, and inclusivity, especially for elderly and disabled travelers. Attention, interest, usage, and emotion drove satisfaction and planning. | Advances smart tourism theory by unifying GenAI, IoT, and NLP; supports inclusive and sustainable tourism frameworks. | Enhances personalization, accessibility, and sustainability; practical use for tourism operators to improve inclusivity and efficiency. | Limited to Thailand; wristband not fully serving blind tourists; future work on NLP/GenAI in Thai language and accessibility tech. |
| 22 | Kim, Kim, Kim and Hailu (2023) | ChatGPT | TAM, trust theory, information adoption, social influence, learning-by-doing | US | 1689 participants (4 survey groups) | Surveys (mediation analysis) + experiments (ANOVA) | Prior experience with ChatGPT strongly predicts adoption; improves idea generation and personalization but lacks real-time accuracy. Negative news has little effect vs. personal use experience. | Extends TAM by showing “learning-by-doing” shapes adoption more than awareness. | Providers should improve reliability, transparency, and error handling; promote positive user stories to build confidence. | Short-term, intention-based; lacks real-use and long-term effects; calls for longitudinal and observational research. |
| 23 | Shi et al. (2024) | ChatGPT | TPB | China | 536 valid travelers | Survey (CB-SEM) | Privacy, accuracy, and overreliance risks reduce attitude, norms, control, and intention to adopt ChatGPT. | Extends TPB by incorporating perceived risks as second-order factors in AI acceptance. | Providers should strengthen accuracy, privacy, and user trust through secure data and training. | Focused on Chinese travelers; single-method; future research should diversify samples, test multiple models (e.g., TAM, UTAUT), and assess broader tourism impacts. |
| 24 | Abou-Shouk et al. (2024) | ChatGPT | Affordance-actualization theory, Ad credibility model | UAE & Oman | 506 travelers (255 UAE + 251 Oman) | Survey (PLS-SEM, MGA) | Relevance, credibility, usefulness, and intelligence foster trust, which increases intention, satisfaction, and loyalty. Privacy/security concerns weaken trust–intention link. | Advances AI adoption by combining affordance-actualization and credibility frameworks; shows dual mediator/moderator roles of trust and privacy. | Emphasize personalized, credible recommendations; strengthen privacy and transparency policies. | Limited to Arab context; demographic moderators not fully tested; future studies should include cross-cultural and qualitative approaches. |
| 25 | Kang et al. (2024) | Gen AI (ChatGPT) | Media richness theory, trust transfer theory, value-based adoption model (VAM) | South Korea | 578 valid participants | Survey (PLS-SEM, MGA) + fsQCA + ANN | Media richness enhances trust in ChatGPT, which transfers to OTAs, increasing perceived value and booking intention. Text + image format most effective. | Integrates media richness, trust transfer, and VAM into a robust framework; highlights perceived value as central to AI adoption. | OTAs should use visuals with text to boost trust; promote transparency and tutorials; integrate GenAI into booking with value emphasis. | Limited to South Korea and ChatGPT; future studies should test cross-cultural and other AI platforms (e.g., Gemini, Claude). |
| 26 | Arora et al. (2025) | ChatGPT | UTAUT2, perceived characteristics of innovation (PCI), flow theory | India | 408 participants | Survey (PLS-SEM) | Performance expectancy, ease of use, hedonic motivation, and trialability drive continuance intention and flow; external influence moderates adoption. | Extends UTAUT2 with PCI and Flow Theory; highlights flow as key to continued use. | Providers should ensure usefulness, ease of use, and engaging design to encourage repeat use. | Limited to Indian millennials; cross-sectional design; future studies should use longitudinal and cross-generational approaches. |
| 27 | Solomovich and Abraham (2024) | ChatGPT | TAM, Big Five personality traits | Israel | 305 tourists | Survey (both PLS-SEM & CB-SEM) | Trust and ease of use predict usefulness and intention; personality traits (extraversion, neuroticism, openness) and age moderate adoption. | Extends TAM by integrating personality and demographic moderators. | Design AI travel tools with attention to user personality and age differences. | Limited to Israel; cross-sectional TAM; needs cross-cultural, longitudinal, and experimental research. |
| 28 | T. Kim et al. (2025b) | Gen AI | Personal innovativeness | South Korea & US | 594 valid respondents | Survey (PLS-SEM, MGA) + fsQCA + ANC | Trust, usage, value, and benefits drive adoption; trust and behavioral intention predict recommendation; innovativeness not a strong moderator. | Extends innovation theory by highlighting trust as primary driver, challenging innovativeness assumptions. | Tourism stakeholders should prioritize building trust and delivering clear utility across user types. | Reliant on self-reported cross-sectional data; limited to South Korea/US; needs longitudinal and wider cross-cultural studies. |
| 29 | Wang et al. (2025) | ChatGPT | S-O-R framework, transmission model of inspiration | China | 254 valid respondents | Survey (PLS-SEM + PROCESS moderation) | Information quality fosters inspiration, which mediates travel planning intention; privacy risk weakens this link. | Advances S-O-R by confirming inspiration as central mediator between AI quality and behavior. | Improve reliability and richness of ChatGPT outputs; mitigate privacy risks for stronger adoption. | China-only; self-reported cross-sectional data; needs cross-cultural, longitudinal, and expanded moderators. |
| 30 | Chauhan and Jishtu (2025) | AI-powered travel planning tools (ChatGPT) | TPB | India | 476 participants | Survey (CB-SEM) | AI anthropomorphism stimulates consumer innovativeness, indirectly shaping attitudes and adoption intentions. Privacy concerns undermine attitudes but do not directly hinder behavioral intentions. | Extends TPB by incorporating domain-specific constructs (anthropomorphism, privacy concerns) that mediate through innovativeness and attitudes. | AI tools should include adaptive personas, co-creative itinerary builders, and privacy reassurances to engage innovative but cautious users. | Cross-sectional data; future research should use longitudinal designs, include constructs such as trust and enjoyment, and test cross-cultural variations. |
| 31 | Wong et al. (2025) | Gen AI (ChatGPT) | Social identity theory | Macau (China SAR) | 448 participants (2 studies) | Experiments (ANCOVA) + mediation/moderation analysis using PROCESS + semi-structured interviews | Human-generated travel plans are perceived as more reliable and more likely to be adopted. Reliability mediates adoption, with travel persona reinforcing trust in human content. | Extends social identity theory by applying it to human vs. machine identity, showing persona alignment drives credibility. | Bloggers/influencers should maintain persona consistency; AI developers should enhance transparency, credibility, and persona-based tailoring. | Focused only on Chinese tourists in Macau; future studies should expand to cross-national and multimedia contexts, and broader persona types. |
| 32 | J. Kim et al. (2025a) | ChatGPT | Information processing paradigm, expectancy–disconfirmation theory | US | 755 participants (4 studies) | Experiments (ANOVA, serial mediation & moderation analysis) + interviews | Fast responses increase usefulness, satisfaction, and travel intentions for simple queries, but slower responses enhance credibility for complex tasks, especially for patient/low rationalism users. | Extends information-processing and expectancy–disconfirmation by identifying speed as a peripheral cue moderated by complexity and user traits. | AI systems should adapt response speed to task complexity, using staged outputs, progress indicators, and user-controlled pacing to balance speed and credibility. | Based on ChatGPT 3.5, US samples, and cross-sectional designs; future research should test other models, cultures, prompt quality, and longitudinal adoption. |
| 33 | Liu et al. (2025) | Gen AI-enabled avatars (ChatGPT 3.5 via API) | Experience economy framework, self-determination theory | US | 559 older adults | Survey (PLS-SEM) | GenAI avatars fulfill autonomy, competence, and relatedness needs, enhancing well-being. Well-being boosts travel confidence, which mediates travel intention. | Extends SDT and experience economy by linking avatar-guided virtual experiences to well-being and confidence as pathways to real-world travel. | Virtual travel should be designed to foster autonomy, competence, and social connectedness for older adults, enhancing well-being and confidence. | US-only sample and cross-sectional design; future studies should adopt longitudinal/experimental designs, test diverse populations, and address ethical concerns (accessibility, privacy). |
| 34 | Kundan et al. (2024) | ChatGPT | TAM, UTAUT | China | 488 valid participants | Survey (PLS-SEM) | Language barriers drive ChatGPT adoption; translation use strengthens travelers’ confidence and approach intentions. | Integrates TAM/UTAUT with language-barrier typology, showing barriers channel into adoption via translation use. | Destinations should integrate ChatGPT-based translation across the travel journey, ensuring offline functionality and privacy-conscious design. | Single-country focus (China) and cross-sectional survey; future research should test diverse destinations, alternative translation tools, and digital literacy as moderators. |
| 35 | Kim, Kim, Kim and Kim (2023) | ChatGPT | Information value, message framing | US | 1522 participants (4 studies) | Experiments (ANOVA, mediation, & moderation analysis) | Ethical/quality issues reduce trust, lowering adoption. Errors trigger algorithm aversion. Moral decoupling occurs when recommendations are concrete (performance judged separately from ethics). | Clarifies trust as a mediator between concerns and adoption, extending persuasion/information-value theories to AI contexts. | Developers must strengthen trust through accuracy, transparency, and safeguards; emphasize ethical assurances for abstract info and precision for concrete info. | US-only online experiments; future studies should test live systems, cross-cultural samples, and longitudinal designs. |
| 36 | Duong et al. (2024) | ChatGPT | Parasocial interaction theory | Vietnam | 606 valid participants | Survey (Hayes PROCESS mediation & moderation analysis) | Parasocial interaction (PSI) enhances satisfaction and continuance intention, with satisfaction partially mediating. Technology anxiety weakens these effects. | Extends PSI theory to AI travel contexts, showing satisfaction as the emotional bridge to continuance and tech anxiety as a boundary condition. | AI tools should use interactive, personalized styles to foster PSI, while reducing anxiety through intuitive design and transparent communication. | Vietnam-only sample and cross-sectional design; future research should adopt longitudinal/cross-cultural approaches and examine roots of tech anxiety. |
| 37 | Han et al. (2024) | ChatGPT | Cognitive–affective–normative (CAN) framework | South Korea | 474 participants | fsQCA | Continuance intention results from multiple driver configurations; informativeness is necessary, combined with hedonic appeal and social norms. | Extends CAN framework by showing continuance arises from holistic configurations, not single drivers. | Providers should ensure informativeness as baseline, while fostering hedonic value and social acceptance. | Limited generalizability due to survey-based, single-country data; future studies should test longitudinal, cross-cultural, and additional constructs (trust, ethics, privacy). |
| 38 | Xu et al. (2025) | ChatGPT | Pleasure–arousal–dominance (PAD) theory | China | 428 valid participants | Survey (PLS-SEM) | Ubiquity enhances dominance, entertainment fuels pleasure, anthropomorphism drives arousal. Pleasure/dominance boost continuance and WOM; arousal mainly promotes advocacy. | Extends PAD theory to AI tourism, linking design features to emotional responses and user behaviors. | Developers should design for accessibility, entertainment, and anthropomorphism to maximize engagement and advocacy. | Chinese sample and cross-sectional design; future work should use cross-cultural, longitudinal, and experimental approaches. |
| 39 | Ali et al. (2023) | ChatGPT | Lay theory, affordance–actualization (A-A) theory | Not specified | 446 participants | Exploratory (video-based perception) + survey (PLS-SEM) | Relevance, credibility, usefulness, and intelligence predict trust; intelligence strongest. Trust drives adoption intentions. | Extends A-A theory and ad credibility by modeling trust as mediator between affordances and adoption. | Providers should emphasize personalization, transparency, and continuous learning to cultivate trust. | Reliance on online panels; limited set of antecedents; future work should test transparency, accuracy, personalization, and cross-cultural variations. |