This appears less like a classification-capability problem than an incentives and enforcement problem. What measurable abuse-response SLA should an ad platform provide?