Evaluating CAI Travel Re commendations: The Effects of AI Hallucinations and Transparency Cues on Perceived Source Credibility and Behavioral Intentions

Sammanfattning

Background: AI hallucinations in conversational artificial intelligence (CAI) are becoming a concern for Online Travel Agencies (OTA), because AI hallucinations can damage consumer trust and disrupt consumers from their booking behavior. Purpose: The purpose of the study is to conduct research on how AI hallucinations in brand-embedded CAI context influence perceived source credibility and behavioral intentions, and if brand familiarity, AI familiarity, or transparency cues can serve as a protective shield to credibility damage. Method: The researchers applied the S-O-R (Stimulus-Organism-Response) framework and conducted a quantitative study to test significant effects. The survey (N = 184) with OTA users mostly from Europe, included a simulation of a Booking.com branded CAI chatbot scenario containing an AI hallucination. The data was analyzed using reliability analysis, Pearson correlation, linear regression and independent sample t-tests in SPSS. Conclusion: The findings suggest that AI hallucinations are the dominant driver of damaging perceived source credibility and brand credibility. Participants who saw the hallucination showed dramatically lower credibility scores. In addition, neither brand familiarity nor AI familiarity served as a protective shield against credibility damage, while transparency cues showed a modest but significant positive effect. Perceived source credibility strongly predicted booking intention. OTA brands, marketing managers, product managers in digital marketing and AI engineers, as well as digital interface designers, can use these findings with regard to chatbot design and construction to focus on AI output accuracy over brand familiarity as a mechanism of trust, and to integrate transparency cues directly into the AI-powered chatbots, to control the damage.