This dataset contains 834 records of cloud seeding activities reported to NOAA in the United States between 2000 and 2025. Each record includes the following fields: year, season, state, seeding agent, deployment apparatus, and the stated purpose of the cloud seeding operation. The dataset was constructed using a multi-stage PDF-to-text extraction pipeline combining PyMuPDF, pytesseract, and Unstract's LLM Whisperer. Extracted text was parsed and structured using OpenAI’s o4-mini language model, which inferred the relevant data fields from semi-structured report content. To assess data quality, we manually reviewed a random sample of 200 records (n = 200). The dataset achieved 94.72% average accuracy across all six fields.