15 Languages, One Indexed: The noindex Strategy We Chose and When We Will Reverse It

WebRecast publishes in 15 languages. Google is allowed to index exactly one

of them. Every non-English page ships a robots meta tag reading

noindex,follow, while the sitemap lists only English URLs.

That looks like sabotage. It is the opposite: on a new domain with no

external links, translated clones are the fastest way to teach Google your

site is a wall of near-duplicates. This is the reasoning, the numbers, and

the exact conditions under which we will flip each language on.

The arithmetic that decided it

The site is a tools collection. Every tool page exists in 14 translations

plus the English original: 20,932 static pages in total, 1,495 URLs in the

sitemap. The translated pages carry the same functionality, the same

structure, and, at the time of the decision, largely the same text with the

words swapped.

Now put that next to the crawl data. Over 30 days the domain earned 11

impressions and 1 click in Search Console. Google was already treating the

site as barely worth a crawl. Multiply the surface by 14 more languages and

every crawl visit gets spent deciding whether a German clone of a page it

declined to index in English deserves indexing. That decision is free for

us to make and expensive to make Google make.

What noindex,follow actually does

The directive is two instructions, and the second one is the point:

• noindex: drop this URL from the index and do not show it.
• follow: still traverse the links on the page.

So the translated pages keep working as internal link surfaces. Link equity

and discovery flow through them. hreflang alternates stay declared, so

Google knows the language cluster exists. If we flip a language on later,

the pages are already crawled, known, and internally linked. Nothing has to

be rebuilt. This is a soft off switch, not a demolition.

The common alternatives are worse here

We considered three other designs and rejected each for measurable reasons.

1. Index everything. Multiplies the near-duplicate surface by 15. On a

domain with zero backlink history, this invites duplicate-content

demotion across the whole cluster rather than per language.

2. Separate hreflang-complete indexing with partial translations.

Mixed-language pages are the worst of both worlds: thin in the target

language, duplicated in structure. Translation coverage across the site

was and is uneven.

3. Separate domains or subdomains per language. Splits an already weak link

profile into fifteen weaker ones. Unthinkable without a real link

foundation.

The English-first option concentrates every signal we earn onto one page

per tool. When the domain is unknown, being the strongest possible version

of one page beats existing faintly in fifteen versions.

Translation work is paused, with conditions to resume

With non-English pages out of the index, producing more translations has

zero search effect, so we stopped. The decision document lists the two

conditions that reopen the work, and both have to be true:

1. English pages are indexing at scale, tracked by a scheduled Search

Console API check that inspects a fixed priority set plus a random

sitemap sample.

2. Search Console impressions arrive from countries matching a specific

language, which is demand evidence for that language and no other.

This turns a taste debate into a measurement. When Japanese impressions

appear in the query data, Japanese gets produced. Not before.

If you run a multilingual site, copy the checklist

1. Count your actual surface: total pages versus sitemap URLs. If the ratio

is high and your domain is young, you have the same decision to make.

2. Check your impressions by country in Search Console before commissioning

any translation. Zero demand means zero return.

3. Use noindex,follow rather than removal for the languages you park. Keep

the internal links and hreflang alive so reversal is cheap.

4. Keep one language fully indexed and fully maintained. Concentration is

the strategy, not deprivation.

5. Automate the reversal condition. A scheduled API check with written

thresholds removes the temptation to guess.

The uncomfortable part of this strategy is that it is slow. Parking

fourteen languages means acknowledging the site earns its languages one at

a time, with evidence. That is the deal.

---

Which languages does your own traffic actually justify today? Query your

country report before answering. If you want the reference version of the

rules involved, the robots meta and hreflang breakdown is here:

https://webrecast.com/en/seo-tools