The 404 pages Google loved too much

The 404 pages Google loved too much

A marketplace noticed thousands of junk URLs creeping into its index — typo'd paths, deleted listings, random garbage. Crawl budget was being eaten alive, and real pages were getting crawled less often.

The data: Search Console showed 14,000 indexed URLs that should not exist. Each one returned a 'Sorry, not found' message — to users.

The investigation checked status codes, not page content. Every one of those dead URLs returned HTTP 200 OK. The single-page app caught unknown routes client-side and rendered a friendly 404 component, but the server happily served the app shell with a 200 for literally any path. Google treated thousands of nonexistent pages as valid, indexable URLs.

Turns out a 'nice' JavaScript 404 with a 200 status is an invitation: Google will index every typo and every deleted listing you ever had.

The fix: make the server return a real 404 status for unknown routes (or 410 for deleted listings), while still rendering the friendly component.

Result: junk URLs fell from 14,000 to under 200 over five weeks, and crawl frequency on real product pages rose noticeably as budget stopped leaking into the void.