Why Is Traffic Growing in Some Languages but Not Others? Check the Data Before Blaming Translation

Published: October 14, 20267 min

How reading tools work

Listen reads the article aloud. Speed read shows phrases in sequence at your chosen pace. Language practice compares available translations. Save keeps a bookmark in this browser; find it in the player’s bookmarks.

Share this article

Share on X (opens in a new tab)Share on Bluesky (opens in a new tab)

A website publishes in twelve languages. Japanese and Korean articles gain readers, while Spanish and French barely move. The tempting explanation is that those markets simply aren't interested.

But a review of one multilingual publishing operation revealed a more surprising possibility: the system choosing which pages to improve was missing some of the struggling pages in the first place. It's hard to rescue a page that your own dashboard has made invisible.

The same site, twelve very different trends

Here are article page views with a search engine recorded as the referrer, comparing October 7–9 with October 11–13, 2026 (Japan time):

Language	Oct 7–9	Oct 11–13
Japanese	26	311
English	1	13
Korean	12	63
Simplified Chinese	2	4
Traditional Chinese	16	35
Spanish	2	2
Brazilian Portuguese	1	0
Indonesian	1	6
Thai	1	4
Vietnamese	4	3
French	1	1
German	0	2

These are search-referral page views, not Google Search Console clicks or search impressions. Some counts are sampling-based estimates, the comparison days differ, and the smallest figures are volatile. They show uneven outcomes, not proof that a particular rewrite succeeded.

First hidden cause: the report cut off the very pages that needed help

The site's search-performance import stopped after 2,000 rows, with an explicit partial-data status. A second report revealed query-and-page combinations absent from the page-level list: 40 for Traditional Chinese, 30 for Vietnamese, 28 for Brazilian Portuguese, 16 for Simplified Chinese, and 13 for Thai. These are query-page rows, not numbers of articles.

Why would weaker languages disappear? Search Console returns its higher-click rows first. If a publisher applies one shared limit across languages, established winners are easier to keep while pages with impressions but few clicks get pushed toward the cutoff.

Then the improvement system sees the winners, congratulates itself on excellent data coverage, and overlooks the pages that need improvement. It's like evaluating a restaurant chain using only receipts from busy branches. Of course every branch in the spreadsheet is busy.

Google's API documentation explicitly says it prioritizes top rows and doesn't guarantee every row. A better approach is to split queries by language, date range, or page group, keep track of partial imports, and never turn a missing row into an assumed zero.

Second cause: a translation was graded as a vocabulary exercise

One Simplified Chinese article was flagged because its title and opening didn't contain English terms from a search about tennis grunting and ball speed.

The reader's question is whether making a sound changes the speed of a shot. A natural Chinese answer may explain that question perfectly without copying an English search phrase. Yet a checker based on exact word overlap can fail it.

The chain looks simple: pick the most visible query, split it into words, check whether those words appear in the headline and introduction, and mark a mismatch if they do not.

Matching characters is not the same as answering a question. This distinction matters even in one language; across twelve, it can reward clumsy keyword stuffing over clear writing. The multilingual editor turns out to be a vocabulary quiz with a fancy dashboard.

Third cause: the headline was fixed while the article answered something else

Consider two searches involving the same search engine. One person wants to learn how to use it. Another wants to get their website discovered by it. An article for site owners can contain every relevant brand name yet still be the wrong answer for the first reader.

Changing only the title or a search preview won't solve a mismatch in the actual question. Nor will automatically lifting the sentence with the greatest keyword overlap from the article.

When the article genuinely answers the query, improve its title and description. When the question is different, rewrite the opening and the relevant body sections so the answer is there.

A practical order for fixing multilingual traffic

Follow each page through five separate checkpoints:

Discovery: Can search engines find and crawl the correct language URL?
Indexing: Is the page eligible and selected for search?
Impressions: Which real questions triggered the page, in which language and country?
Clicks: Does the displayed promise match both the question and the article?
Delivery: Was the edited version published, recrawled, and then measured?

Start by recovering missing search data for every language. Replace word-presence tests with checks that compare a reader's question to the article's actual answer. Separate small headline fixes from full editorial repairs. Then compare performance under matching conditions.

Market size, competition, and local habits still matter. Equal growth in all twelve languages cannot be promised. But a page excluded from the measurement pipeline has not failed a fair test of demand.

Before telling the quiet branches to work harder, check whether their sales receipts ever reached headquarters.

References (4)

The case figures come from an anonymized operational review recorded on October 14, 2026. No causal share of the traffic growth was attributed to individual changes.