Microsoft removed more than 3 billion ads and banned nearly 400,000 websites from its network in 2021 -- up from the 1.6 billion and 270,000 sites it took down or banned in 2020.
The company also suspended nearly 270,000 accounts from its network, attributing the increase in monitoring to artificial intelligence such as zero-shot learning to identify new patterns and methods with malicious intent. The technology identifies mischievous activities in text, images and video.
“Microsoft Advertising witnessed an increased number of threats and varied attack patterns in 2021,” Neha Garg, principal product manager at Microsoft, wrote in a post. “In our commitment towards offering our users safe and trusted experiences, we invested in product, algorithm, and process-based enhancement and improvements.”
Garg wrote about continued malicious advertising behaviors in 2021.
While there were many new patterns, fake celebrity-endorsed investment ads was one area that showed growth.
Microsoft took measures to avoid promoting these ads. Machine learning, anomaly detection models, and fraud investigation experts’ insights helped to suspend nearly 10,000 accounts and take down more than 200,000 of these types of ads.
Microsoft also increased investments in multilingual models to detect non-compliant content in more than 30 languages, and made advancements in the way humans moderate the content to capture more insights from reviews. Intelligent tools allow human reviewers to more easily identify links between various accounts and discover various fraud rings.
Microsoft also addressed customer complaints to remove low-quality ads. In 2021, the company received about 70,000 complaints related to ads not being compliant per its advertising policies.
After investigating each complaint, the company found about 60% were in violation, most related to trademark infringements.
Complaints ranged from trademark infringement -- at about 34,700 -- to policy violations, at about 20,934.
A few complaints were related to unlicensed gambling sites, phishing, unauthorized government-service provider websites, and other user safety concerns. In response to complaints, nearly 650,000 ads in violation were removed from the network.