Harnessing AI for good: world-leading trial monitoring alcohol ads

Today, we're publishing the results of a cutting-edge trial that used artificial intelligence (AI) at scale to check whether alcohol ads are sticking to the rules. The findings show that the overwhelming majority of alcohol advertisers are compliant, and the trial itself marks a major step forward in how innovative technologies like AI can strengthen regulation and support responsible advertising.

Our goal was to see whether AI could help us spot potential problems that traditional monitoring or public complaints might miss, while giving a clearer picture of overall compliance. We've been using AI since 2023 to help us proactively monitor the online advertising landscape, but this is the first time we've used it to check such a large volume of ads against a whole section of the advertising rules.

Using our AI-based Active Ad Monitoring System, we analysed almost 6,000 paid-for ads shown to the UK public in early 2025, spanning alcoholic drinks, alcohol-free alternatives and related promotions across search, display and social media.

Each ad, including its text and imagery, was assessed by state-of-the-art large language models (LLMs) - advanced AI tools that can interpret and apply our rules. Ads flagged by the LLMs as being at risk of breaking the rules were then reviewed by our human experts to determine whether the content flagged was indeed a likely breach. This process combined the scale and speed of AI with expert human judgement to ensure accuracy.

Overall, our findings are encouraging. The overwhelming majority of alcohol advertisers are following the rules:

• Around 96% of the ads we reviewed were likely to comply with the alcohol specific advertising rules.
• Just 1-3% appeared to break them.
• The remaining 1% needed further review.

The most frequent issues we observed across the small number of ads that appeared to break the rules included: misleading or unauthorised health or nutritional claims, ads promoting irresponsible drinking, and content likely to have particular appeal to children. All these issues were seen across ads by both large national hospitality chains and smaller advertisers, including independent venues.

Advertising by the alcohol-free sector raised more consistent concerns - around 48% of alcohol-free product ads were flagged for potential breaches. Almost all of these related to unclear or missing ABV labelling, an issue which can be easily rectified but an important requirement that matters for people who avoid alcohol altogether, including for health, religious or personal reasons.

We'll now be contacting the advertisers responsible for the clear breaches we identified and publishing advice for the alcohol-free sector, particularly on ABV labelling.

Ultimately, this trial shows how we're using AI to deliver world-leading regulation, and the central role it now plays in supporting key areas of our work. The lessons from this trial will guide future improvements, making our AI tool even more precise and useful across different areas.

By combining AI with expert human judgement, we get a clearer picture of where ads meet or fall short of the rules, enabling faster and more effective action on emerging risks and helping keep UK advertising responsible and trusted.

Adam Davison, Director of Data Science at the Advertising Standards Authority, said:

"We believe AI has the potential to make our online ad regulation much faster and more effective. This report shows that the latest advances in AI can be used to help our experts scan large volumes of ads and check them against lots of complex rules more quickly. It highlights that the alcohol sector is generally sticking to the rules, which is great news. We've also spotted places where we think we can do more, and deliver better regulation thanks to our rapid adoption of AI."

Read the pulse report on alcohol advertising.