PageImages and Page Content Service select the "most representative" image in the article, based on various conditions (mostly position and image size). If the images in the article don't look appropriate, they won't select anything. The selected image is used in all kinds of software features (it's what we use in our task cards, but also it's used as a hero image in the apps, it's shown in search results, in the popups shown when hovering a link on desktop, in the VisualEditor link dialog etc); generally an article with a page image is a much nicer user experience than one without.
Maybe it would make sense to align our image selection logic with the PI/PCS criteria so if some of the images are appropriate as page images and others aren't, we prefer the appropriate ones.
This isn't really relevant for the main goal of the Add an image feature (a retention / onboarding mechnism for new users), but inasmuch as we see improving article content a side goal, it might be beneficial for that.