This study compares traditional and large language model (LLM)-based search for image geolocation, determining the location where an image was captured. Sixty participants were randomly assigned to use either traditional or LLM-based search engines to assist in geolocation tasks. Results showed that participants using traditional search outperformed those using LLM-based search in terms of accuracy. Traditional search users tended to issue shorter, more keyword-based queries and added more terms when reformulating, while LLM-based search users issued longer, more natural language queries but had shorter search sessions and consistently rephrased their initial queries. Qualitative findings revealed that LLM-based search users struggled with query formulation, often failing to communicate their intent effectively to the search engine. Participants using LLM-based search also faced challenges in formulating queries in different languages and in understanding the search results. The study highlights the importance of query formulation in image geolocation and the need for better understanding of how users interact with LLMs. The results suggest that traditional search engines may be more effective for image geolocation tasks, particularly for non-expert users. However, LLM-based search has the potential to provide more conversational and natural language interactions. The study also emphasizes the need for further research into how users form mental models of LLMs and how to improve the design of LLM interfaces to better support user interactions. Overall, the study provides valuable insights into the differences in strategies and user behaviors when using traditional versus LLM-based search for image geolocation.This study compares traditional and large language model (LLM)-based search for image geolocation, determining the location where an image was captured. Sixty participants were randomly assigned to use either traditional or LLM-based search engines to assist in geolocation tasks. Results showed that participants using traditional search outperformed those using LLM-based search in terms of accuracy. Traditional search users tended to issue shorter, more keyword-based queries and added more terms when reformulating, while LLM-based search users issued longer, more natural language queries but had shorter search sessions and consistently rephrased their initial queries. Qualitative findings revealed that LLM-based search users struggled with query formulation, often failing to communicate their intent effectively to the search engine. Participants using LLM-based search also faced challenges in formulating queries in different languages and in understanding the search results. The study highlights the importance of query formulation in image geolocation and the need for better understanding of how users interact with LLMs. The results suggest that traditional search engines may be more effective for image geolocation tasks, particularly for non-expert users. However, LLM-based search has the potential to provide more conversational and natural language interactions. The study also emphasizes the need for further research into how users form mental models of LLMs and how to improve the design of LLM interfaces to better support user interactions. Overall, the study provides valuable insights into the differences in strategies and user behaviors when using traditional versus LLM-based search for image geolocation.