Freigeben über

Microsoft Monetize - Ad Quality models

  • Artikel

Ad quality is a cornerstone of your business, and at Microsoft Monetize, we provide technology to help you manage it effectively. The integration of the Xandr platform into Microsoft Advertising introduces two AI-powered ad quality models that identify and filter low-quality or harmful content. These models use large language models (LLMs) and neural networks to detect and filter ads containing misleading, inappropriate, or low-quality content.

Enable Ad Quality model categories

These ad quality models are available through the Categories in your ad quality rules. You can block these two categories at both the network and publisher levels, allowing for tailored ad quality management.

For detailed instructions on applying categories to your ad quality rules, see the following resources:

Categories

Currently, ad quality models support two categories that leverage AI and deep learning to enhance ad quality management:

  • Clickbait model (Category 695)
  • Image sensitivity & aesthetics model (Category 703)

Note

These models will evolve to support more categories over time. Microsoft Monetize continuously enhances ad quality filtering, and additional topics may be added to these models in the future.

Clickbait model (category 695)

Clickbait is a form of online advertising that uses sensational, exaggerated, or misleading headlines, images, or phrases to attract attention and entice users to click on a link. These ads often lead to:

  • Low-quality or irrelevant content
  • Malicious or fraudulent websites
  • Decreased platform credibility

The Clickbait model applies AI-driven content analysis across various ad categories, including:

  • Finance
  • Cryptocurrency
  • Health
  • Gaming

The model uses advanced language analysis to identify patterns commonly associated with deceptive or misleading ad content. This model uses sophisticated LLMs to detect and filter misleading ad content, ensuring high-quality ad experiences.

Image sensitivity & aesthetics model (category 703)

This model checks ads for sensitive topics and visual aesthetics to improve user experience. It scans for content related to:

  • Pharmaceuticals
  • Adult content
  • Weapons
  • End-of-life services
  • Weight loss
  • Religion
  • Drugs
  • Smoking
  • Cryptocurrency