‘Could we know whether you had good time there?’ Real and artificial tourist pictures underlying (un)happiness through AI and LLMs

Purpose. The aim of this paper is to demonstrate the extent to which it is possible to automatically extract photo content data in terms of human-embedded emotions such as un/happiness and if it is possible to analytically detect and recreate tourists' emotions through Artificial Intelligence (AI) and Large Language Models (LLMs). Methodology. We built and trained a classifier machine to automatically analyse pictures of real tourists in five international destinations. We then compared them with a silicon sample of tourists generated by AI and LLMs. Findings. This study extends our knowledge and practice of tourists' emotions identification by utilising new methods involving AI. This study demonstrates the advantages of combining facial analysis of individuals' emotional reactions, by offering an alternative to the self-reports. Also, our work provides further evidence of the limited adequateness of LLMs in associating the information in the provided textual data (prompts). Originality. To the best of our knowledge, this is the first empirical study that attempts to analytically detect and recreate tourists' emotions through Artificial Intelligence (AI) and large language models (LLMs) by utilising synthetic data based on a silicon sample of tourists. Practical implications. Our work shows how to evaluate consumers' psychological state or emotional reactions towards a specific destination with new tools not completely relying on tourists' self-reports. In our work, instead, we discuss a new possible approach to capture real emotions and how they could be measured, allowing extracting virtual insights into tourists' experiences and preferences towards specific destinations and locations.