In terms of harmful advertisementsGoogle has had to evolve its defense strategy and relies on the capabilities of l’IA Gemini to automate detection and blocking. This approach made it possible to intercept more than 99% of non-compliant ads before they were served.
A balance sheet for 2025 is the blocking or elimination of a record number of 8.3 billion adsthe suspension of 24.9 million accounts, including 602 million advertisements and 4 million accounts associated with scams.
In the European Union alone, Google has removed more than 1.6 billion ads and suspended 2 million ad accounts.
Gemini understands the intention behind an ad
Unlike previous systems, which relied on analyzing specific terms, Gemini analyzes hundreds of billions of signals, including account age, behavioral cues and campaign patterns.
This contextual analysis helps identify malicious content designed to evade traditional filters, and the technology allows for near-instantaneous intervention.
At the end of last year, the majority of responsive ads created in Google Ads were reviewed in real time, blocking harmful content as soon as it was submitted.
When a threat manages to slip through the cracks, Gemini has helped process four times more user reports compared to the previous year.
It’s also a battle between AI
With the use of Gemini, Google now favors more targeted and granular blocking of individual advertisements, rather than mass suspensions of accounts. By distinguishing ” more finely a credible offer of a sophisticated lure “, AI has enabled an 80% reduction in incorrect advertiser suspensions in 2025.
That being said, the malicious actors are themselves using generative AI to create deceptive content on an unprecedented scale, using obfuscation techniques and URL redirections to appear legitimate. The challenge for Google is to always stay one step ahead.
