Gambling Industry Political Engagement, 2020–2026: A Proof-of-Concept for Auditing of Political Transparency Data using a GPT-assisted Pipeline

Background: Commercial political activity is increasingly recognised within the broader framework of commercial determinants of health. However, UK gambling-related political engagement is fragmented across separate transparency systems. This study examined whether a GPT-assisted, independently reconciled workflow could integrate these records and provide a more complete picture of disclosed political activity. Methods: A retrospective analysis was conducted of UK gambling-related political engagement from 1 January 2020 to 25 August 2026 across three streams: (1) Electoral Commission political donations; (2) MPs’ declared hospitality/gifts and funded visits; and (3) ministerial meetings within the Department for Digital, Culture, Media and Sport / Department for Culture, Media and Sport (DCMS). GPT was used to retrieve, structure, link and classify records against a reference frame of gambling organisations, which was derived from multiple public sources, including Gambling Commission registers, Companies House records, company and trade-body information, and historical ownership and corporate-relationship evidence. After classification, each of the three streams underwent an independent GPT reconciliation pass, followed by bounded human adjudication of unresolved cases. A final audit compared the final donations and hospitality datasets with previous UK journalism and civil-society reporting. Results: The principal analysis identified 127 donations with confirmed active commercial gambling connections, worth £7.73 million, from donors with a confirmed material commercial gambling connection active on the donation date. Of this value, 85.4% went to the Conservative political family and 13.0% to Reform UK. MPs declared 176 qualifying hospitality/gift events involving 71 MPs, worth £236,162, plus four funded visits. The ministerial-meeting census contained 272 gambling-related events, of which 86 involved principal commercial-industry actors. Across the 138-group company universe, meeting frequency was positively associated with both donation frequency (Spearman ρ=.390; p=.00004) and hospitality frequency (ρ=.487; p=.00001), although these associations are observational and non-causal. In the final audit against previously reported figures, 291 of the observed donation/hospitality records had no apparent or obvious match in the prior-reporting examined; and one genuine omission in this analysis was identified and corrected (related to hospitality). Conclusions: Systematic linkage across transparency systems revealed a scale and pattern of gambling-connected political activity that is not apparent from individual registers or prior headline reporting. GPT made this integration tractable, but independent reconciliation and human adjudication were essential. Integrated monitoring of political activity may therefore be important for understanding the commercial determinants of health in gambling and other harmful commodity industries.