GraphRAG is a card game that uses graph theory to improve generative AI chatbots, specifically Retrieval-Augmented Generation (RAG). The game simulates how a chatbot would answer questions about a topic by first spinning a spinner, which determines the level of information to provide. If it lands on STANDARD, the player answers the question directly; if RAG, they fetch relevant documents and use them to inform their answer; and if GRAPH RAG, they select a relationship type and draw all connected clue cards in one step. The game board represents a graph database, with clue cards as vertices and relationships between them as edges. GraphRAG improves chatbot performance by reducing hallucinations, improving explainability, and enabling answers about unseen topics.