Dataset for 'Using AI agents to assemble population-level data for visibility and animal welfare insights: a case study of greyhounds racing in the UK'.

This deposit comprises the pseudonymised population-level dataset assembled for the manuscript "Using AI agents to assemble population-level data for visibility and animal welfare insights: a case study of greyhounds racing in the UK" (Cobb & Coghlan, 2026: https://doi.org/10.3389/fanim.2026.1868726). It describes greyhounds racing under Greyhound Board of Great Britain (GBGB) regulation in the UK between 1 January 2022 and 31 March 2026 (inclusive), assembled by cross-referencing five publicly accessible online sources: the Irish Coursing Club (ICC) Studbook, the GBGB racing database, the British Greyhound Stud Book, Greyhound-Data.com, and Greyhoundstats.co.uk. Contents. A dog-level table of 31,030 individual greyhounds (demographics, career metrics, country-of-origin classification); a race-level table of 1,267,124 dog-starts across 22 licensed tracks; an origin-resolution table documenting the strict process used to classify country of origin; a reference dictionary mapping GBGB remarks codes to plain-language meaning and category; a SQLite mirror of all tables for convenience; a data dictionary; supplementary figure and table data; and the Python pipeline and analysis code. Privacy and pseudonymisation. The deposit contains no human personal identifiers. Trainer, owner, and breeder names sourced from the public registries during data assembly were removed prior to deposit. Dog names, registered greyhound names, sire names, and dam names were also removed; each dog is represented by a sequential pseudonymised identifier (D00001 – D31030). The encrypted mapping key linking pseudonymised identifiers to source registry names is retained securely by the corresponding author for verification purposes and is not transferable. AI tool disclosure. The data assembly and analysis code were authored by M. L. Cobb, using Claude (Anthropic, Inc.) operating in agent mode for defined retrieval, parsing and structuring tasks, under continuous human supervision. The corresponding author reviewed all code, validated all outputs, and is responsible for the integrity of the methodology and results. AI tools were not used for data interpretation or manuscript generation. See manuscript Sections 3.2 and Ethical Considerations for full details. Licence and citation. The dataset is released under a Creative Commons Attribution 4.0 International (CC BY 4.0) licence; the accompanying analysis and assembly code are released under the MIT licence. Reuse of either should cite this Zenodo record and the accompanying manuscript. Reproducibility. Methods Sections 3.1 and 3.2 of the manuscript provide sufficient detail for independent replication of this deposit from the same public sources. The deposit is structured to align with FAIR principles for scientific data management. Subsequent re-assemblies of the data will be linked to the same Zenodo concept DOI to enable longitudinal comparison.