Abstract: Generative Artificial Intelligence personalized recommendations play a critical role in tourism information search. How to leverage GAI personalized recommendations to reshape the information search experience and enhance decision-making efficiency has become an urgent issue for the tourism industry. Grounded in information search-related theories and the 5A model, this study systematically investigates the effects of different types of GAI personalized recommendations namely proactive versus reactive recommendations, recommendation with high versus low adaptive algorithms and recommendations provided by GAI alone versus human-GAI collaboration on tourists’ information search behaviors and decision-making across three pre-travel stages: travel need recognition, itinerary planning, and booking decision-making.
The theoretical contributions of this study are structured as follows. First, based on the core characteristics of users’ information search behaviors across different pre-travel stages, this study proposes a stage-based classification of GAI-driven personalized recommendations. Most existing studies treat the pre-travel stage as a single holistic phase, overlooking heterogeneity in users’ needs, motivations, and decision tasks across stages. This study therefore adopts the 5A model and refines the pre-travel stage into three sequential phases: travel need recognition, itinerary planning, and booking decision-making. These stages exhibit a progressive relationship, in which need recognition provides foundational information for itinerary planning, and itinerary planning, in turn, enables booking decisions. Accordingly, this study proposes stage-specific types of GAI personalized recommendations and emphasizes the matching relationship between recommendation types and stage-specific search tasks and informational needs, thereby extending the application of the 5A model in tourism information search research.
Second, this study develops a theoretical framework explaining how different types of GAI personalized recommendations influence tourism information search behaviors and decision-making across the three pre-travel stages, thereby enriching and advancing the theoretical system of tourism information search research. The dynamic interaction mechanisms of GAI, heterogeneous user characteristics, and diverse tourism contexts introduce uncertainty into the effects and pathways through which personalized recommendations influence decision behaviors. By systematically examining how different recommendation types operate across pre-travel stages, this study fills a gap in the literature on GAI-enabled tourism information search and highlights the importance of understanding GAI personalization from an information search theory perspective.
Third, this study reveals the underlying mechanisms and boundary conditions through which different types of GAI personalized recommendations influence users’ psychological processes and decision-making behaviors across pre-travel stages, thereby advancing research on the integration of personalized tourism marketing and AI technologies. Existing studies primarily focus on the overall effectiveness or single-dimensional attributes of personalized recommendations, with limited attention to the heterogeneous mechanisms of different recommendation types. By incorporating stage-specific task characteristics, such as cognitive load, perceived risk, emotional engagement, and information needs, this study explains how different GAI-driven recommendation types affect decision-making through psychological mechanisms including creativity, trust, and perceived information comprehensiveness. This contributes to a more comprehensive understanding of the integration between GAI and personalized marketing and addresses gaps in the classification and mechanism analysis of GAI recommendation types.
This study also offers significant practical implications. Although GAI is increasingly integrated into tourism information search and personalized recommendation systems, tourism enterprises still face challenges such as overly uniform recommendation strategies, inaccurate user need identification, and poor alignment between recommendation content and search tasks. By classifying GAI recommendations according to pre-travel stages and emphasizing the matching between recommendation types and task characteristics, this study provides actionable insights for optimizing information search experiences and improving recommendation efficiency.
Specifically, during the travel need recognition stage, users are more sensitive to the attractiveness and inspirational value of information. Enterprises should therefore adopt more creative and exploratory recommendation strategies to stimulate travel interest and demand. During the itinerary planning stage, users focus more on information integration, comparative evaluation, and comprehensibility; thus, recommendations should emphasize interpretability and comparability of travel plans. During the booking decision stage, users prioritize trustworthiness, transparency, and risk controllability. At this point, high-reliability human-GAI collaborative recommendations can enhance decision confidence and improve conversion outcomes. However, for standardized, low-risk products such as attraction tickets, autonomous GAI recommendations may provide superior user experiences due to faster response and smoother interaction.
Finally, this study provides practical guidance for the design of GAI-based tourism recommendation systems. The findings demonstrate that different recommendation types influence user psychology and decision-making through mechanisms. Therefore, system developers should enhance the capability of identifying users’ information search stages and dynamically adjust recommendation logic and information presentation accordingly. In addition, recommendation systems should strengthen process transparency and interpretability by presenting recommendation rationales, reference information, generative logic, and key constraints. This can improve users’ understanding of recommendations and reduce concerns regarding AI hallucinations and algorithmic “black-box” effects. Furthermore, incorporating human-GAI collaborative recommendation and process visualization can enhance user trust and further improve the effectiveness of GAI personalized recommendation systems in tourism contexts.

Key words: generative artificial intelligence, personalized recommendation types, tourism information decision- making, Information Processing Theory, 5A model