Digital humanities research has revolutionized the study of ancient inscriptions by providing researchers with access to immense epigraphic corpora. However, traditional search methods for these databases rely primarily on exact or fuzzy keyword matching, limiting researchers’ ability to find semantically related inscriptions. This paper presents a new approach to searching ancient inscriptions using vector embeddings and semantic similarity, implemented through a hybrid search system that combines semantic search with keyword matching and large language model re-ranking. Our system processes Greek and Latin inscriptions from the Ancient Graffiti Project database, embedding them in a high-dimensional vector space that captures semantic meaning beyond exact text matches. Our process is designed for reproducibility, using open data and code, and shows promise in preliminary evaluation. Our results demonstrate the system’s capability to identify thematically related inscriptions that would be missed by traditional search methods, offering new possibilities for epigraphic research and discovery.