<b>Streaming SSR was sending Google a half-finished page</b>

<b>Streaming SSR was sending Google a half-finished page</b>

A media site adopted React streaming SSR for faster perceived loads. Three weeks later, article body content started disappearing from Google's index — titles ranked, but the text Google had on file was just the intro paragraph.

Streaming sends HTML in chunks: shell first, then Suspense boundaries resolve and flush as data arrives. The article body sat behind a Suspense boundary fed by a CMS call that took 3–6 seconds. For real users this was fine — the page filled in progressively.

But Googlebot's render has a finite patience window. On slower CMS responses, the crawler captured the rendered DOM after the shell flushed but before the body Suspense boundary resolved. It indexed the skeleton: headline, byline, an empty article container.

We moved the article body out of the slow Suspense boundary and resolved its data before the first byte. The shell now ships with the body already inside it.

—Pages with full body indexed: 60% → 99%
—Organic to long-form articles: +44% in 9 weeks

Streaming is great for humans who wait. The crawler captures one frame — make sure the words are already in it.