<b>The pages Google saw once and never updated</b>

<b>The pages Google saw once and never updated</b>

An ecommerce site refreshed prices and stock daily. But Google's cached versions were weeks stale — showing sold-out items as available and old prices, tanking click quality and triggering merchant warnings.

The site was a SPA with soft navigations. After the initial server render, every subsequent navigation happened client-side via the router — no new full document, no fresh HTML for the server to hand back. When Googlebot recrawled a product URL, it got the same SSR shell from a long-lived cache, then had to render JS to get current data. But the SSR snapshot itself was cached at build time with stale prices baked in, and the crawler often trusted the rendered shell.

The real issue: the SSR output was generated once at deploy and cached indefinitely, so "server-rendered" content was frozen at the last build.

We switched product pages to per-request SSR with a short cache, so each crawl got live data in the first HTML.

—Pages with current price in index: 30% → 97%
—Merchant data-quality warnings: 140 → 6

Server-rendered isn't the same as fresh. Static SSR is just a screenshot from build day.