Komparativní hodnocení personalizovaných turistických itinerářů vytvořených modely generativní AI

This bachelor's thesis compares personalized tourist itineraries generated by selected generative artificial intelligence models. The aim of the thesis is to evaluate the quality of itineraries generated by GPT-5.5, Claude Sonnet 4.6, and Gemini 3.1 Pro in terms of factual accuracy, logical and temporal feasibility, and the degree to which they satisfy user preferences. The thesis also proposes an original evaluation methodology based on previous research, combining both objective and subjective evaluation criteria. The research was conducted on a dataset of 27 tourist itineraries generated from nine input prompts representing different combinations of traveler types and preferences. The results showed that all tested models achieved a high level of factual accuracy as well as logical and temporal feasibility. The greatest differences were observed in fulfilling user preferences, particularly with regard to budget requirements and the incorporation of individual traveler preferences. Gemini 3.1 Pro achieved the best overall performance. The thesis also highlights the identified limitations of generative artificial intelligence in creating personalized tourist itineraries and proposes directions for the further development of evaluation methods in this area.