When a main image only lazy-loads via script and alt is empty, agents get nothing citable and assistive tech receives no content. A crawl-friendly approach puts the main image and descriptive text into the initial HTML in a readable form.
Have templates emit srcset and alt directly, with alt describing the product, colour or key spec rather than stuffing keywords; inline critical styles and defer the rest so the image and key text remain in raw HTML with scripts disabled. Follow the accessibility quick reference for clear structure so different readers reach the same facts.
This improves readability and accessibility only; it does not alter the product or issue a verdict on whether a single crawl parsed successfully.