The product images existed. Google Images swore they didn't.
A furniture retailer drove 22% of revenue from Google Images. After a frontend rebuild, image search traffic dropped 67% in a month. The product pages looked identical. The photos were right there.
The symptom was specific: pages still ranked in web search, but image impressions cratered. That narrowed it.
Turns out the rebuild moved every image to a homegrown lazy-loader using IntersectionObserver, with the real URL in data-src and a 1x1 transparent placeholder in src. Below-the-fold images only swapped src when scrolled into view. Googlebot renders a tall viewport but doesn't scroll like a user — most product shots never entered the observer's range, so their src stayed the transparent pixel. Google indexed a 1x1 gif as the product image.
We swapped the custom loader for native loading='lazy', which keeps the real URL in src and lets the browser (and crawler) handle deferral. Real src, real image, day one.
Image search traffic recovered to -8% of baseline in five weeks, then +14% by week ten as fresh shots got recrawled. Image-driven revenue came back to 19%.
Native lazy-loading isn't just simpler. It's the only version Googlebot reliably sees.
Hydrate Diaries
@HydrateDiaries
The product images existed. Google Images swore they didn't.
Этот пост опубликован в Telegram-канале Hydrate Diaries. Подписаться можно по ссылке: @HydrateDiaries.