China is one of the countries most severely affected by flood disasters worldwide. Long-term and systematic flood impact data are essential for flood risk assessment, flood impact model calibration, disaster prevention and reduction. However, an open, long-term, China-focused event-scale database with detailed flood disaster impact indicators remains lacking. Here we present the China Flood Disaster Impact Database (CFDID), which contains 1716 flood disaster event records from 1994 to 2023, including 1391 top-level (L1) event records. CFDID is compiled from 66 official documents and covers five flood-related disaster categories: rainstorm flood, typhoon, snowmelt flood, ice-jam flood, and dam-break flood. Each record includes date, location, disaster category, eight impact indicators, and data source. The database was constructed using a workflow that combines optical character recognition (OCR), large language model (LLM)-based information extraction, and manual verification, thereby improving the efficiency of converting unstructured text information into structured event records while ensuring data quality. Based on these L1 records, CFDID achieves total coverage ratios of 67.22 % for deaths and missing persons, 72.25 % for affected population, 60.70 % for affected cropland area, 65.68 % for collapsed dwellings, and 74.86 % for direct economic losses relative to official annual flood disaster impact totals during 1994–2023. Compared with commonly used international flood impact databases, CFDID contains more flood disaster event records and has higher coverage for most impact indicators. CFDID also includes agricultural impact indicators, which are rarely available in international hazard impact databases. CFDID is available at https://doi.org/10.5281/zenodo.21699973 and will be continuously updated by year.