2
2 Comments

A soft-404 bug was quietly telling Google my Arabic pages existed, when they were actually just showing English content

Nine days ago I posted here about getting a solo project indexed on
Google, Bing, and Yandex. Since then I found something that undid
part of that work without me noticing.

My site supports 5 languages. I assumed "translated" meant translated.
It didn't.

The bug: 22 tool pages had translated content in some fields but not
others, and instead of 404ing on the missing locale, the page would
silently render with English content, while still returning a 200
status and telling search engines "this is the Arabic version of this
page" via hreflang.

That's a soft-404. It looks completely fine to a human visitor. To
a crawler, it's a mismatch between what a page claims to be and what
it actually contains, and enough of those can make the engine stop
trusting your whole multi-language cluster, not just the broken pages.

I found 88 affected URLs (22 tools × 4 non-English locales).

While auditing this, I also found something bigger: only 9 out of
112 tools
had their actual page titles translated. The other 103
fell back to English regardless of language selected. Since launch.
Never flagged, because it doesn't throw an error, it just silently
under-delivers.

The fix, for anyone hitting the same thing:

  • Every locale × page combination needs a single source of truth
    for "does a real translation exist here", not separate hardcoded
    lists in your switcher, your hreflang generator, and your static
    build config.
  • Untranslated locales should 404, not silently fall back to a
    different language's content under a URL that claims otherwise.
  • Your 404 page itself needs noindex and zero hreflang tags, since
    it has no real cross-language equivalent.

None of this shows up in a normal "check your SEO" checklist. It only
shows up if you specifically go looking for the gap between what your
switcher says exists and what your build actually generates.

Fixed 15 of the 22 tools so far (translated content into all 4
languages). The other 7 aren't shipped yet, so I left them out
entirely rather than fake it.

on August 23, 2026
  1. 1

    The strongest part is the gap between what the site claims exists and what the crawler actually receives. A page can look “working” while quietly undermining an entire localization cluster.