Business

Google Says AI Stopped 99% of Bad Ads Before Anyone Saw Them

By CM Chaney · April 17, 2026

Google blocked more than 8.3 billion ads last year. More than 99% of them were caught before a single user ever saw them.

That is the headline figure from Google’s 2025 Ads Safety Report, released Thursday, which credits the company’s Gemini AI models with a fundamental shift in how online ad enforcement works — moving from reactive takedowns to proactive blocking at the point of submission.

The scale is significant. Eight-point-three billion removed ads represents a 63% increase from the 5.1 billion removed in 2024. Among those, 602 million ads and 4 million advertiser accounts were specifically tied to scams. In the U.S. alone, Google removed 1.7 billion ads and suspended 3.3 million accounts.

What makes the report notable is not just the volume — it’s the mechanism. Google says its Gemini models now analyze hundreds of billions of signals per ad, including account age, behavioral patterns, and campaign characteristics, to identify malicious content before it ever enters the ad system. Unlike earlier keyword-based detection, Gemini is designed to understand intent — catching scam ads even when they’re constructed to evade traditional filters.

“Our systems caught over 99% of policy-violating ads before they ever served,” Google’s VP of Ads Privacy and Safety Keerat Sharma told journalists. “Our goal is to stop badness before it is ever exposed to any user.”

Gemini also changed how Google handles user complaints. The company says it processed four times as many user reports in 2025 compared to the year prior, using AI to triage and act on flagged content faster than human teams alone could manage. At the same time, Google says the system reduced incorrect advertiser suspensions by 80% — a figure it is using to argue that tighter enforcement doesn’t mean more collateral damage for legitimate businesses.

The context matters. Scammers are increasingly using generative AI to create convincing fake ads at industrial scale — impersonating brands, fabricating celebrity endorsements, and deploying deepfakes. Google’s TechCrunch interview noted the company blocked fewer advertiser accounts overall even as it blocked more ads — a deliberate strategic shift toward targeting individual bad ads rather than broader account suspensions.

The report is partly a statement of capability and partly a competitive signal. Google processes the world’s largest volume of digital advertising, and its ability to position Gemini as a consumer safety tool carries weight beyond the ad industry. Whether the 99% threshold holds as scammers adapt their own AI tools is the open question. Google said it plans to expand instant ad review to more formats in 2026.