April 16, 2026

AI Helped Google Block Millions of "Malicious Ads"

Google stated this Thursday that tools using Artificial Intelligence (AI) Gemini helped it thwart about 8.2 billion online ads that violated its policies last year. AFP "We have truly immersed ourselves in the task of rethinking our systems using AI," Keerat Sharma, Google's vice president of advertising privacy and security, told reporters. "Our latest models better understand our intent, which helps us detect malicious content and block it preventively, even when it has been designed to evade detection," he highlighted. An Annual Advertising Security Report, published by Google, indicated that the internet giant's systems intercepted more than 99% of ads that violated its rules before they were shown to online users. "Our goal is to stop bad practices before they are exposed to any user," said Sharma. Advertising system tools, powered by Gemini, analyzed hundreds of billions of "signals"—such as advertiser account age, behavior, and campaign patterns—to determine if anomalies existed, he detailed. "We took a step back and thoroughly rethought how we wanted to build advertising security, adopting a perspective starting from the ground up," commented Sharma. "Thanks to Gemini, we have been able to take that distance and combine hundreds of billions of signals to achieve a much deeper understanding of an advertiser's true intent, based on the sum of all the information we have about them," he noted. "The combination of these elements gives us a highly nuanced perspective on whether a specific ad's intent is positive or potentially harmful." When malicious actors resort to generative AI to create misleading ads on a large scale, Gemini helps detect and block them in real time, Google reported. Among the advertising tactics prohibited by Google's policies are "deepfakes," which is the improper use of people's images to make them appear real. Another layer of defense in Google's advertising system is advertiser verification. "Verification is a tool we use to ensure that the advertiser is, in fact, a legitimate entity," explained Sharma.