How does Magnet scrape images from article pages?

Magnet processes the item properties in the below order to find the image that will be used as thumbnail:

1.<meta itemprop="thumbnailUrl" content="<image url>"/>

2.<meta itemprop="image" content="<image url>"/>

3.<meta property="og:image" content="<image url>"/>

If there is no image associated with the content, widgets will show the image that is associated with the company-instance (the URL is configurable in Magnet's Publishers' dashboard).