Google Ads Safety Report: AI helps to slow down the wave of fraud
In 2024, Google blocked over 5 billion non-compliant ads and suspended more than 39 million ad accounts—many of them before the first ad was even served. According to the new Ads Safety Report, this was made possible in particular by advances in large language models (LLMs).
With the annually published "Ads Safety Report", Google takes stock of its measures to ensure a safe advertising environment. The 2024 edition shows: Advances in artificial intelligence - especially in the large LLMs - enable the company to detect and prevent fraudulent content at an early stage.
"Our LLMs now require only a fraction of the data of previous models to identify abuse patterns and distinguish between legitimate companies and fraud attempts," the report states. Over 50 improvements to these models were introduced in 2024.
These technical developments led to more than 5.1 billion "bad" ads being blocked or removed and 39.2 million accounts being blocked - a large proportion of them in the onboarding process before the first ad went live.
In 2024, particular attention was paid to fraudulent ads that used generative AI to spread fake statements from public figures. "A dedicated team of over 100 experts developed new countermeasures, such as updating our misrepresentation policy," writes Google. As a result, over 700,000 advertising accounts were permanently blocked. The number of corresponding scam reports has fallen by 90 percent.
The report also shows the impact of AI-supported systems on publishers. 97 percent of the pages against which Google took measures in 2024 were recognized by automated models. A total of 1.3 billion pages were checked and excluded if necessary.
The report also highlights the expansion of advertising transparency in the election environment: in 2024, over 8,900 new election advertisers were verified worldwide and more than 10 million ads were removed from unverified accounts. According to Google, it was the first company to introduce transparency obligations for AI-generated content in election advertising.
The report thus provides an example of how AI can contribute not only to increasing efficiency, but also to quality assurance in digital advertising ecosystems. "The threat situation is constantly changing - our investments in technology and governance are geared towards ensuring a secure framework for users and advertisers," concludes the report.
The full report is available on the Website from Google available.
