Google's "2025 Advertising Security Report" released on Thursday local time showed that the company blocked a record 8.3 billion ads globally last year, a significant increase from 5.1 billion the previous year. However, the number of suspended advertiser accounts was significantly lower, which triggered attention from the outside world about changes in its platform governance methods.
Google attributes this departure from the trend of "blocking more ads but banning fewer accounts" to its deeper adoption of a series of artificial intelligence models, including Gemini, in the field of advertising security, so that it can identify and block illegal advertising earlier and more accurately.
Google said that with the help of AI systems represented by Gemini, it has achieved greater automation and refinement in blocking illegal content during the ad review process. In 2025, more than 99% of illegal ads identified by the system will have been blocked before being displayed to users. The report pointed out that on the one hand, this reflects the platform’s enhanced ability to “pre-manage” illegal content. On the other hand, it also means that the entire platform’s law enforcement focus is shifting from “one-size-fits-all” bans on advertisers to “click-to-kill” targeting specific advertising creatives themselves.
Google also pointed out that the surge in the number of blocked ads is also related to the abuse of generative AI by criminals. More and more scammers are using generative models to create highly misleading false advertising content on a large scale, forcing the platform to identify risks in larger and more complex patterns. Google's internal Gemini model is used to find common patterns in these large-scale advertising campaigns, locate problems in advance, and intercept them at an earlier stage. This law enforcement logic also echoes Google's overall strategy of fully and deeply embedding Gemini in core products such as search, workspace, and advertising. From advertising campaign generation and strategy optimization to violation detection and real-time risk control, the degree of AI participation continues to increase.
Judging from the specific types of violations, fraud-related content is particularly prominent among blocked advertisements and suspended accounts. Google said that violations related to fraudulent activities in 2025 involved 602 million ads and 4 million advertiser accounts. In the United States alone, Google removed more than 1.7 billion ads and suspended 3.3 million advertiser accounts in one year, with ad network abuse, misleading promotions and sexual content being the most common violations. In the Indian market, which has the largest user base, Google blocked 483.7 million ads in 2025, almost double the previous year, but the number of account suspensions dropped from 2.9 million in 2024 to 1.7 million. The most prominent local violations include trademark abuse, financial services-related violations, and copyright issues.
At an online media communication meeting, Keerat Sharma, vice president and general manager of Google Advertising Privacy and Security, said that the company has consciously transformed over the past year towards "more targeted and more AI-based law enforcement." He emphasized that compared with the previous "blunt weapon" of blocking advertiser accounts, the current strategy is more refined interception at the "creative level" and a more granular approach to specific advertising materials. Sharma said this shift will help Google reduce the proportion of incorrectly suspended accounts by 80% in 2025 compared with the same period last year.
Sharma also pointed out that the "layered defense" mechanism deployed by Google on the front end of ad delivery - such as requiring advertisers to complete identity verification and prove the authenticity of the company and business before delivery - has played an important role in preventing malicious actors from creating accounts, which is one of the reasons for the decline in the overall number of account suspensions. In his view, as Google continues to introduce new defense measures and criminals continue to adjust their countermeasures, these data will still fluctuate in the future, but the company's overall goal is to block harmful content as early as possible in the advertising life cycle.
Overall, this annual report shows the latest adjustments to Google's advertising security strategy driven by AI technology: on the one hand, the platform uses more powerful models to deal with the "amplification effect" brought by generative AI. On the other hand, the platform shifts from "punishing advertisers" to more refined "eliminating bad ads" in law enforcement, trying to find a new balance between protecting user safety, maintaining trust in the advertising ecosystem, and reducing accidental injuries to compliant advertisers.