Imagine searching for "sports race" and getting a jumble of horse races, Formula 1, and even esports. Frustrating, right? A new research paper explores how AI can help refine image searches by suggesting related queries that narrow down results to specific visual themes. This innovative approach, called cross-modal query suggestion, analyzes the initial search results and identifies visually consistent clusters. For example, if you search for "sports race," the system might suggest related queries like "horse race," "car race," or "bicycle race," allowing you to quickly zero in on the type of race you're looking for. The researchers introduce CroQS, a new benchmark dataset designed to test and compare different cross-modal query suggestion methods. CroQS consists of initial queries, their corresponding image search results, and human-generated suggested queries that better capture the different visual themes within those results. Initial experiments show promising results, with AI-generated suggestions significantly improving the accuracy and relevance of image searches. However, there's still room for improvement compared to human-generated suggestions. This research paves the way for more intuitive and effective image search tools, making it easier to find precisely the image you're looking for. Future research will likely focus on improving the understanding of user intent and context to deliver even more personalized and relevant search suggestions.