Google reports improved ad safety as Gemini tools detect policy-violating ads earlier

Singapore – Google says new tools powered by its generative AI model Gemini have strengthened the company’s ability to detect and prevent harmful advertisements, according to a blog post published by Keerat Sharma, Vice President and General Manager for Ads Privacy and Safety at the company.

In the post, Sharma said Google’s safety teams continue to combat increasingly sophisticated advertising threats from bad actors who attempt to distribute scams and other malicious ads online.

“Our safety teams work around the clock to stop bad actors that use increasingly sophisticated, malicious ads. In 2025, Gemini-powered tools dramatically improved our ability to detect and stop bad ads: Our systems caught over 99% of policy-violating ads before they ever served, and we’re continuing to evolve our defenses to stay ahead of even the most advanced schemes,” he said.

According to Google, Gemini-enabled systems analyse hundreds of billions of signals—including advertiser account age, behaviour patterns, and campaign activity—to identify potential threats before ads are displayed. The company said the newer models focus more on understanding intent rather than relying primarily on keyword detection, enabling them to detect attempts to circumvent policies.

Google reported that in 2025 it blocked or removed more than 8.3 billion ads and suspended 24.9 million advertiser accounts. This included 602 million ads and 4 million accounts linked to scams.

The company also cited its advertiser verification program as an additional safeguard designed to prevent fraudulent advertisers from entering its ecosystem by validating identities and strengthening trust in ads shown to users.

Google said the rise of generative AI has enabled scammers to create misleading ads at scale, but the same technology is also helping the company detect such content more quickly. Gemini-based systems can review many ads automatically at the submission stage. By the end of 2025, Google said most Responsive Search Ads created through Google Ads were reviewed instantly, allowing harmful content to be blocked before publication.

The system also processes user reports more efficiently, enabling Google to act on more than four times as many user complaints in 2025 compared with the previous year.

Sharma added that improved detection accuracy has helped reduce incorrect advertiser suspensions.

“By analyzing beyond images and text patterns, Gemini can better distinguish between a credible offer and an intricate lure — a level of nuance that helped us reduce incorrect advertiser suspensions by 80% last year. This accuracy allows us to focus on two things in tandem: prioritizing the removal of harmful content while helping honest businesses keep their ads running.”

Google said it plans to expand instant ad reviews powered by Gemini to additional advertising formats in the future as part of its ongoing efforts to maintain ad safety while supporting legitimate advertisers.