Bots or Humans? The Role of AI and eWOM in Travel Decisions Amid Crisis

Published: 17 November 2025| Version 2 | DOI: 10.17632/wmn8sht2tj.2
Contributor:
Musa Zakaria

Description

Guided by Theory of Planned Behavior (TPB) and Source Credibility Theory (SCT), the study addressed RQ1: How does crisis communication source (CCS: AI chatbot vs. eWOM) affect visit intentions (VI) to paradox destinations (risky yet attractive places, e.g., politically unstable with cultural appeal) via perceived source credibility (PSC) and attitudes toward the destination (ATD), moderated by perceived coping efficacy (PCE)? RQ2 examined differences between conditions. Hypotheses: H1: CCS influences PSC (AI > eWOM). H2: PSC positively affects ATD. H3: ATD positively affects VI. H4: Serial mediation (CCS → PSC → ATD → VI). H5a–c: PCE moderates paths (stronger at high PCE). H6a–b: Paths stronger in AI vs. eWOM condition. Data Description and Gathering Quantitative experimental survey data from 352 Chinese adults (63.6% female, mostly 25–34 years old, bachelor's degree holders, occasional travelers). Collected via Credamo.cc in 2025 using convenience sampling. Participants randomly assigned to eWOM (n=179) or AI chatbot (n=173) conditions, viewing simulated crisis info for fictional destination "Solara" (political instability + attractions). Measures: PSC (5 items), ATD (3), VI (3), PCE (4) on 7-point Likert scales; demographics as covariates. Cross-sectional; suitable for causal tests but limited generalizability. Interpret: Higher scores = stronger perceptions; use means/SDs for descriptives, β/p-values for paths. Usage: Replicate in SPSS/AMOS for mediation/moderation; apply to tourism strategies. Notable Findings Analyzed via SPSS/AMOS (ANCOVA, PROCESS Model 6, CFA, MGA). Reliability α >.89; model fit good (CFI=.96, RMSEA=.04). AI more credible than eWOM (p=.008; H1 supported). PSC → ATD (p<.001; H2) and ATD → VI (p<.001; H3) positive. Serial mediation significant (b=0.1572, CI [0.0354, 0.2986]; H4 supported). PCE moderated PSC → ATD (p=.010) and ATD → VI (stronger for low-PCE; H5b–c supported, H5a not). MGA: Paths stronger in AI condition (e.g., PSC → ATD: β=.90 vs. .79; p=.01; H6 supported). Data Interpretation Data shows AI chatbots enhance credibility over eWOM in paradox contexts, driving attitudes and intentions via mediation. Low-PCE tourists more influenced by credible sources for risk mitigation; high-PCE more resilient. Limitations: Fictional scenario, Chinese sample, self-reports. Implications: Use AI for factual crisis comms, eWOM for emotional ties; hybrid approaches for recovery in unstable destinations.

Files

Categories

Social Sciences, Tourism, Marketing, Behavioral Psychology, Travel Behavior, Crisis Communication

Licence