What changed
Amazon has introduced a new AI evaluation system using Vision-Language Models (VLMs) to assess the quality of A+ Content on its Marketplace. This system evaluates the complete content experience, providing actionable recommendations for brands to enhance their product detail pages.
What this means in practice
The new approach allows brands to transform average content into high-quality content that is more coherent, relevant, and useful. By rendering the full product detail page as a single snapshot, the VLM evaluates the content as customers see it, considering how different assets work together. This holistic evaluation helps brands make informed purchasing decisions and build lasting relationships with customers, leading to improved revenue for selling partners.
Limitations and availability
The system uses a multi-model fusion architecture, combining VLM semantic scores with OCR readability signals in a lightweight CatBoost classifier trained on expert annotations. It also employs SHAP to explain quality decisions and generate actionable improvement recommendations. The end-to-end pipeline is implemented on Amazon SageMaker AI, from data ingestion and annotation through model hosting, explainability, and production monitoring. The evaluation system is designed to meet the requirements of scale, accuracy, and explainability, with scores matching expert quality standards and results explaining what to fix.
