Scammers work fast, but Google says its AI assistant Gemini works faster. New data shows it is stopping billions of “bad ads” before people even see them.
As reports show that online scams are getting worse and more prolific, work to protect users from them is more important than ever, says Google vice president and general manager, ads privacy and safety Keerat Sharma.
According to Google’s 2025 Ad Safety Report, new Gemini models allowed Google to block or remove over 8.3 billion ads, 99% of which before they were even seen by anyone.
Google also suspended 24.9 million accounts last year.
Here in Aotearoa, this looked like 225 million ads removed and 45,000 advertiser accounts suspended. The top five violations were for sexual content, dating and companionship, abusing the ad network, inappropriate content and misrepresentation.
The top violation globally was abusing the ad network which saw over 1.29 billion ads blocked or removed. According to Google, this means the ad tried to trick or circumvent its ad review processes, including links to compromised sites or malware, or manipulated images or text.
North Star
Blocking harmful content before it is served is Google’s North Star, says Sharma.
It is Google’s belief that the internet should be free and open, but achieving that requires a high degree of trust from everyone involved, including users to brands, he adds.
Sharma describes the three ways Google is using Gemini to improve ad safety:
- Blocking ads with greater precision
- Working faster than scammers
- Stopping threats while supporting business.
Stopping threats before they start
Sharma says the models used have evolved from keyword-based systems to analysing hundreds of billions of signals, including account age, behavioural cues and campaign patterns.
The models better understand intent and therefore are better at detecting and blocking malicious content before it reaches people, he adds.
In 2025, Google blocked or removed over 8.3 billion ads and suspended 24.9 million accounts.
The second layer is Google’s advertiser verification programme, which validates identities and aims to stop bad actors before they enter the system.
Scams are fast, Gemini is faster
Gemini’s ability to review ads instantly means harmful content is being detected and blocked upon submission. It’s a capability Google is looking to bring to more ad formats this year, says Sharma.
The tool’s speed also means it can process user feedback more efficiently. This helps the team respond when threats manage to slip through.
Finally, Gemini can now better distinguish between credible and fake offers. This extra layer of nuance allows it to better analyse the intent of the ad, with the aim of differentiating honest businesses from the scam artists.
The outcome? An 80% reduction in incorrect advertiser suspensions in 2025, says Sharma.
Looking forward
As scams and bad actors continue to evolve and change, Google will also innovate to stay ahead, says Sharma.
While many scams target older – typically less tech-savvy – generations, Google’s ad safety programme is focused on everyone. There are no special protections for certain users, he says.
Looking ahead, Google will investigate how else Gemini can be made more integral to its infrastructure.
Accountability and transparency are also important so users can know who is behind the ads they are seeing and have an opportunity to learn more about the company, says Sharma.
The post Google’s Gemini blocked over 8.3 billion ads in 2025 – report appeared first on stoppress.co.nz.
Discussion about this post