Google on Thursday said its tools powered by Gemini artificial intelligence helped block around 8.2 billion online advertisements that violated its policies last year.
“We really rethought how to improve our systems using AI,” said Keerat Sharma, Google’s vice-president for ads privacy and safety, during a media briefing, reports AFP/BSS.
“Our latest models have a stronger ability to understand intent, enabling us to identify malicious content and block it in advance, even when it is designed to evade detection,” he added.
According to Google’s annual Ads Safety Report, more than 99 per cent of policy-violating ads were detected and removed before being shown to users.
“Our goal is to stop harmful content before it ever reaches users,” Sharma said.
He noted that Google introduced 35 updates to its advertising policies last year to keep pace with evolving scam tactics.
Sharma explained that Gemini-powered systems analyse hundreds of billions of signals — including advertiser account age, behavioural patterns, and campaign activity — to identify suspicious content.
“We took a step back and rebuilt our ad safety approach from the ground up,” he said.
“With Gemini, we can combine vast amounts of data to better understand an advertiser’s intent based on everything we know about them,” Sharma added.
“This allows us to form a more nuanced judgement about whether an ad is legitimate or potentially harmful.”
Google also acknowledged that bad actors are increasingly using generative AI to create misleading ads at scale, but said Gemini enables real-time detection and blocking of such content.
Prohibited tactics under Google’s policies include deepfakes of public figures and misuse of authentic images.
Sharma added that advertiser verification remains a key defence, with more than 90 per cent of ads on Google coming from verified entities.
“Verification helps ensure that advertisers are legitimate,” he said, adding that the company continues to strengthen its systems to counter increasingly sophisticated threats.
