The Image Extraction tab configures how the system chooses a representative image for each page. That image is used in search results and in agent citations.
The crawler first checks standard metadata (e.g. og:image). If none is present or you want to override it, you can use Image URL Patterns and/or Image XPaths (under Advanced). → Open Crawler Settings (Image Extraction tab)
When enabled, the system automatically picks a representative image for pages that don’t have suitable og:image (or similar) metadata. It runs after each crawl: it learns global images to exclude (logos, icons), then uses heuristics to pick the best content image per page.
URL substrings that identify product or content images (e.g. /images/products/, /uploads/). The first image whose URL contains any of these patterns is used as the page thumbnail. This is more flexible than XPaths and also considers background-image styles.
XPath expressions used to find image URLs on pages that lack standard og:image (or similar) metadata. They are evaluated in the order you list them. Image URL Patterns are tried before XPaths.
//img[@id='main_image']/@src//div[@class='article-image']/img/@srcog:image. Add one XPath per row and reorder as needed.