Utilizing knowledge graphs to verify AI-generated content is an important approach to alleviate hallucination issue of large language models (LLMs). Previous work has typically focused on the impact of integrating knowledge graph-based verification to the outcomes produced by LLMs, while neglecting to assess the effectiveness and reliability of the verification process itself. In this paper, we present KGFV , a benchmark specifically developed to comprehensively evaluate the procedure of knowledge graph-based fact verification. Specifically, KGFV aims to provide an end-to-end evaluation framework for assessing the verification capabilities involved in comparing and validating AIGC with knowledge graph information. To this end, KGFV features a diverse set of factual scenarios, encompassing simple facts, complex multi-hop reasoning, comparative analyses, and set operations. Furthermore, KGFV also provides a wide range of intermediate information, which can serve as a robust foundation for future research. Experiments with several state-of-the-art fact verification approaches on KGFV demonstrate that there is still a long way to go to the effective and reliable fact verification between AIGC and knowledge graph.