This report presents the findings of a major ASA trial exploring how artificial intelligence (AI) can strengthen the regulation of online advertising. Using our in-house Active Ad Monitoring system, we analysed nearly 6,000 paid-for alcohol-related ads served to UK audiences in early 2025, across social media, search and display. 

The trial tested whether large language models (LLMs) – advanced AI tools capable of analysing both text and imagery – could assess ads against the alcohol-specific rules in Section 18 of the CAP Code. Each ad was screened by AI and any potentially non-compliant content was reviewed by ASA experts. This approach combined the scale and speed of AI with expert human judgement. 

Overall compliance was very strong. Around 96% of ads were likely to meet the rules, with just 1–3% appearing to break them and a small proportion requiring further review. Most of the likely breaches related to unauthorised health or nutritional claims, irresponsible drinking messages, or content likely to appeal to children.