Google Utilizes AI to Suspend Over 39 Million Ad Accounts for Suspected Fraud

Google announced on Wednesday that it had taken a major step in fighting ad fraud by suspending a staggering 39.2 million advertiser accounts on its platform in 2024. This number is more than triple what was reported the previous year, showcasing Google's intensified efforts to clean up its ad ecosystem.

By harnessing the power of large language models (LLMs) and analyzing signals like business impersonation and suspicious payment details, Google claims it can now suspend the "vast majority" of these ad accounts before they even have a chance to display an ad. It's like catching the bad guys before they can even step into the bank!

Last year alone, Google rolled out over 50 LLM enhancements to bolster its safety enforcement across all its platforms. "While these AI models are incredibly important and have led to significant improvements, we still rely on human oversight throughout the process," explained Alex Rodriguez, Google's General Manager for Ads Safety, during a virtual media roundtable.

Rodriguez shared that Google assembled a team of over 100 experts from various divisions, including Ads Safety, Trust and Safety, and researchers from DeepMind. This team worked tirelessly to analyze deepfake ad scams, especially those involving impersonations of public figures, and developed effective countermeasures.

The company introduced over 30 updates to its ads and publisher policies last year, along with technical countermeasures. These efforts led to the suspension of more than 700,000 offending advertising accounts and a dramatic 90% reduction in reported deepfake ads, according to Google's own claims.

In the U.S., Google suspended 39.2 million advertiser accounts and removed 1.8 billion ads last year. The key violations included ad network abuse, trademark misuse, healthcare claims, personalized ads, and misrepresentation. Meanwhile, in India, which boasts the world's largest population and the second-largest internet market after China, Google suspended 2.9 million accounts and removed 247.4 million ads. The top policy violations in India were related to financial services, trademark misuse, ad network abuse, personalized ads, and gambling and games.

Out of all the suspensions, Google specifically targeted 5 million accounts for scam-related violations, and overall, it removed nearly half a billion scam-related ads.

In 2024, with half the world's population heading to the polls, Google verified more than 8,900 new election advertisers and removed 10.7 million election ads. However, Rodriguez noted that the volume of election ads is relatively small compared to Google's overall ad numbers, so it won't significantly affect the company's safety metrics this year.

Last year, Google blocked a total of 5.1 billion ads and removed 1.3 billion pages. In 2023, these numbers were even higher, with over 5.5 billion ads blocked and action taken against 2.1 billion publisher pages. Google told TechCrunch that the decline in these numbers reflects improvements in its prevention efforts, as better early detection and suspension of malicious accounts means fewer harmful ads make it to the platform.

Additionally, Google restricted 9.1 billion ads last year. However, large-scale suspensions can raise concerns about fairness in applying rules. To address this, Google offers an appeal process that includes human reviews to ensure appropriate action was taken.

"Sometimes, our messaging wasn't as clear or transparent about the specifics, the rationale, or the reasoning behind our actions, which left advertisers confused. We've updated many of our policies and transparency capabilities to better communicate what and why to help advertisers understand our decisions. This has been a major focus for our team in 2024 and will continue into 2025," Rodriguez concluded.