The Authenticity of ChatGPT’s Responses in the Tourism and Hospitality Sector: An Explorative Study on Human Perception of Artificial Intelligence

The study explores the attitudes of tourism/hospitality/gastronomy customers toward AI-generated textual content. Three model scenarios (themes: tourism information center, hotel, restaurant) were outlined as input for the ChatGPT AI model, which generated text responses. Using ChatGPT responses, a text elicitation technique was performed during semi-standardized interviews with regular tourists and hospitality customers. Results showed negative attitudes toward AI-generated textual content to be approximately 1.5 more frequent than positive content. Topics such as noting the high professionalism of the employee (i.e., blinded AI), general satisfaction with the answers, and emphasizing the positive character of the employee emerged for the positive-coded statements. Negative-coded statements were dissatisfaction with short phrases, including greetings/addressing/endings (effort of AI to appear human-like resulted in inauthenticity and strangeness), dissatisfaction due to too long, general or evasive answers (showing high susceptibility in agents), dissatisfaction with too learned, standardized, intrusive, strange, ridiculous or poetic answers. It is concluded that agents could not recognize the responses as AI-generated despite frequently manifesting suspicion and unpleasant feelings. Based on the positive–negative ratio of the statements, interviewed agents were classified into more introverted AI-Accepters (willing to involve AI in services), more extroverted AI-Refusers (not willing to involve AI in services), and AI-Selectors (preferring AI in tourism, but not in hospitality or gastronomy). ChatGPT did not show success when addressing cultural differences in its answers.