For most of its history the ASA waited for complaints. In gambling it no longer does. Many of the gambling rulings now published started with the regulator's own monitoring, and much of that monitoring is done by AI.
What the system does
The ASA's Active Ad Monitoring system captures online ads and uses AI models to flag those that may break the rules, which ASA experts then review. In gambling, the ASA says it is monitoring more than 10,000 online paid ads from UK-licensed operators every month, using AI tools built on large language models to check each one against the full set of gambling rules in the CAP Code and compare it with past rulings. Gambling specialists review potential breaches case by case.
During the 2026 World Cup, which started on 11 June, it scanned more than 13,000 gambling campaigns in under a month.
The system also looks at faces. In its Snapshot Report of 23 July 2026, the ASA said its AI age-prediction model flags people who may be under 30, and reviewers then assess whether they appear under 25. The Midnite ruling of 16 September 2026 shows it at work: an affiliate ad with an AI-generated character, found by monitoring, not a complaint.
Lesson: assume every paid gambling ad you run is read by the regulator's AI, not only by your audience.
Moving into organic content
Paid ads were only the start. In its letter of 28 May 2026 to the gambling APPG and Peers for Gambling Reform, the ASA reported on a review of operators' social media posts between August 2025 and March 2026. Active Ad Monitoring flagged more than 700 posts. Of a representative sample of 388 that experts assessed, about 10% were borderline and about 5% were clearly problematic. The rest were compliant or not ads at all.
A week later, on 4 June 2026, the Gambling Commission and CAP announced an AI-powered content marketing sweep from 11 June 2026, using Active Ad Monitoring and partnerships with social media platforms, alongside a CAP enforcement notice on content marketing.
By 30 September 2026 the ASA reported on about 10,000 pieces of gambling social content reviewed over June and July: 45 ads had strong appeal to under-18s, about 99.5% compliance, and all eight advertisers removed the flagged ads.
What the numbers tell operators
- Coverage is close to total. A post does not need a complaint to reach a reviewer.
- Compliance is high, so misses stand out. When 99.5% of content passes, the remaining posts are the ones that get looked at.
- Borderline cases become rulings. The ASA said it would investigate borderline issues where a ruling helps set the line, which is how individual operators end up as precedents.
- Clear-cut problems are referred on. Posts outside the ASA's remit can be referred to the Gambling Commission, which matters because an ad breach is a licence breach.
The same scrutiny reaches search ads. Our review of paid digital ad rulings includes a gambling search ad that appeared against a self-exclusion query, where the ASA expected extremely robust controls.
Matching the regulator's speed
A regulator checking every ad with AI changes the maths for sign-off. If the check on the other side is a person reading posts at the end of a long day, the regulator will see more than you do. The answer is not to remove people from sign-off. It is to give them the same first pass the regulator has.
- Screen every asset, paid and organic, before it goes live, not a sample.
- Check faces and apparent age as deliberately as copy.
- Check against the precedent rulings, not just the code text.
- Keep a record of each check, so you can show what was reviewed.
Giving your team the same first pass
Voxly Vision checks images, video and copy against the CAP and BCAP Codes, the Gambling Commission's marketing rules and your own brand guidelines before anything goes live, and flags issues for a person to review. A human always makes the final sign-off decision, and every check is recorded, so you can show what was reviewed and when. Diageo used it to cut ad sign-off from 3+ days to about an hour, and review time per asset from 5-10 minutes to about 60 seconds.
See how it works.
Voxly Vision





























