A multilingual publisher can collapse every problem into one dangerously vague sentence:
“The foreign-language side is not growing.”
That diagnosis is almost useless.
Are people failing to discover the pages? Are search results appearing but not getting clicks? Did a social API accept a post that nobody ever saw? Did readers reach an article but have no obvious next article to open? Or is the measurement itself missing?
All of those can produce a zero.
But they are not the same zero.
Think of a multilingual site as a building. Overseas acquisition is the problem of getting people to the front door. Internal circulation is the problem of guiding people from one room to another after they enter.
Complaining about hallway circulation when the front door does not exist is pointless. Bringing one hundred tour buses to a building whose lobby is a wall is equally pointless.
Separating those two systems makes foreign-language “mystery zeros” much easier to diagnose.
0. The short version: do not treat zero as one category
There are two funnels.
External acquisition
publish → discoverable through search or social → actually shown to people → clicked → article loads → visit is recorded
Internal circulation
arrive at an article → read enough to continue → see a next-article option → find it interesting → click → next article loads
They connect, but they are not interchangeable.
If external arrivals are zero, there is no denominator for internal circulation. If internal circulation is zero, the acquisition path may still be healthy.
A dashboard that only says “0” asks one number to do far too much work.
Zero is an outcome, not a root-cause label.
1. In a real audit, identical “foreign zero” values had different causes
In a late-September 2026 operational snapshot from a 12-language publisher, several locales had hundreds of published articles but very different distribution conditions.
| Locale | Published inventory | Social condition | External article views in a 7-day snapshot |
|---|---|---|---|
| Simplified Chinese | 450+ | public posts existed | 0 |
| Traditional Chinese | 440+ | at least one public route existed | 13 |
| Vietnamese | about 420 | no active social route | 0 |
| French | about 430 | no active social route | 0 |
| German | 380+ | public posts existed | 0 |
Vietnamese and French therefore did not have a simple “we posted and nobody cared” problem. At least on social, the route itself was not operating.
Simplified Chinese and German were different. Public posts existed, so the next questions were:
Did the post appear in anybody’s feed? Did anybody care enough to click? Did the link resolve correctly? Did the article load? Did analytics record it?
A social API saying “accepted” does not mean a human saw a single pixel.
API acceptance is not applause. It is not a visit either.
2. Break overseas acquisition into search and social before diagnosing content
Search has its own stages:
URL exists → search engine discovers it → crawler fetches it → page is indexed → result is served → reader clicks
Google explicitly notes that even a compliant page is not guaranteed to be crawled, indexed, or served in results.[1]
For multilingual sites, the relationship among localized versions also matters. Google documents ways to declare localized alternatives through HTML, HTTP headers, or sitemaps.[2]
So “we have 450 translated articles” proves inventory, not acquisition.
It may be a warehouse with no road.
Social is a different pipeline:
copy generated → API accepted → publicly visible → shown in feeds or discovery surfaces → clicked → article reached
A platform can accept a post while an app or account remains non-public. Even when public, low impressions, weak copy, audience mismatch, bad links, or measurement failures can still block traffic.
One green “posted successfully” lamp should not certify the entire delivery network.
3. Even if people arrive, a missing hallway ends the session after one page
Now consider internal circulation.
In the same seven-day snapshot, Japanese pages recorded 92 external arrivals and 52 internal arrivals, while foreign-language pages recorded 48 external arrivals and only 1 internal arrival.
That is not a clean per-user “circulation rate.” These were destination-oriented page observations, not a perfect reconstruction of each person’s reading path.
Still, the gap is a strong warning that foreign-language pages were rarely being reached through internal navigation.
Another audit recorded 2,838 foreign-language pages, with 271 pages showing zero related articles and 82 showing only one or two.
The user experience becomes:
“Welcome!” “Here is your article!” “You finished!” “…and now please leave the premises.”
The hotel has 2,838 rooms.
The hallway stops halfway through.
4. “Show six related articles” is not the end of the job
Internal circulation has several gates:
candidate can be generated → it exists in production HTML → it enters the reader’s viewport → its value is understandable → it matches the reader’s next curiosity → it can be clicked → it reaches the correct article in the same language
A recommendation at the very bottom does not exist for readers who never reach the bottom.
A title-only card may not explain why the next page is worth opening.
A list that only says “more from this category” may fail once the reader feels finished with that category.
Sometimes the better navigation is not “another related article” but a next direction:
- more first-hand experience
- understand the mechanism
- compare alternatives
- see a counterexample
- switch to something lighter
- jump to something completely different but interesting
Six cards labeled “Related” are less human than:
“If that interested you, is this the next thing you are curious about?”
Google’s own link guidance says internal links help both people and Google understand a site and discover other pages, and recommends that important pages receive links from other pages on the site.[3]
Internal links are an SEO mechanism, but first they are simply hallways.
5. The most dangerous misdiagnosis: calling “nobody arrived” “nobody continued”
Suppose Simplified Chinese, Vietnamese, French, or German records zero article arrivals during an observation window.
You cannot yet conclude:
“Foreign-language readers do not continue to a second page.”
Nobody arrived.
Zero visitors entered and zero visitors continued.
That is not a 0% continuation rate.
It is not measured.
Do not solve 0 divided by 0 with confidence in a marketing meeting.
Likewise, three public social posts and zero site visits do not prove zero demand.
Impressions might be zero. Impressions might exist but clicks might be zero. Clicks might occur but the link might fail. The visit might happen while tracking fails.
“Demand is absent” should be a late-stage hypothesis, not the first convenient trash bin.
6. You need origin-to-destination measurement to understand circulation
Destination counts are not enough.
Useful internal-navigation measurement should distinguish at least:
- source article
- source locale
- recommendation shown
- recommendation actually viewed
- recommendation clicked
- destination article
- destination locale
- meaningful reading
- session continuity
- human versus bot or test traffic
For multilingual sites, the critical question is:
Did a reader of a foreign-language article continue to another article in that same language?
One click from a Japanese page to a French page and one natural French-to-French continuation may both look like “internal +1,” but they mean very different things.
Bad measurement can make a busy site look haunted.
Counting bots as readers can make the haunted house look sold out.
Neither is useful.
7. Give each zero a reason code
Instead of showing raw zeroes by locale, classify them.
| State | Meaning |
|---|---|
| ROUTE_MISSING | no major search/social acquisition route is operating |
| PUBLISHED_NOT_PUBLIC | post processing succeeded but the post is not publicly visible |
| PUBLIC_NOT_EXPOSED | public status confirmed, exposure not observed |
| EXPOSED_NOT_CLICKED | exposure observed, no click |
| CLICKED_NOT_ARRIVED | click occurred, arrival not confirmed |
| ARRIVED_NOT_CONTINUED | article arrival occurred, no next-article movement |
| INTERNAL_CANDIDATE_MISSING | too few next-article candidates |
| INTERNAL_VISIBLE_NOT_CLICKED | candidates were visible but not clicked |
| NOT_MEASURED | required instrumentation is missing |
| TOO_SMALL_TO_JUDGE | sample is too small for interpretation |
Now “French = 0” and “German = 0” stop being the same problem.
One may have no road. Another may have a road with no traffic. A third may get visitors who become lost inside the building.
Give the zero an address.
Then the repair owner becomes obvious.
8. Repair order: front door → arrival proof → hallway → next curiosity
The order matters.
1. Build an entrance where no entrance exists
If a social route is inactive, make it publicly reachable before optimizing social copy.
For search, verify locale URLs, crawlability, indexing, and localized-version relationships.
2. Where an entrance exists, split exposure → click → arrival
If public posts exist but site visits are zero, identify the exact drop before multiplying posting volume.
3. Where arrivals exist, verify internal candidates and production rendering
Recommendation logic in source code is worth nothing if the production article shows none.
4. When candidates are visible, recommend directions, not only same-topic pages
Use article-end and context-sensitive transitions to surface the next useful curiosity.
5. Only then blame topic, title, translation, or audience fit
If the page is discoverable, exposed, clickable, reachable, readable, and still weak, content-market fit becomes a stronger hypothesis.
Reverse that order and you end up holding a typography meeting for a road that has not opened yet.
9. Conclusion: overseas acquisition is the front door; internal circulation is the hallway
The multilingual problem was never one vague problem called “foreign languages are weak.”
There were locales with no entrance. Locales with public posts but unknown exposure. Locales that received arrivals but almost no internal continuation. Pages with too few related candidates. Pages where navigation existed in logic but not necessarily in production. Segments that had not yet been measured properly.
Hundreds of articles without an entrance are a warehouse. An entrance without a hallway is a one-room tour. A hallway that always recommends the same room produces “I already saw that.”
The useful dashboard is not merely page views by locale.
It is:
locale × discovery × exposure × click × arrival × reading × next-item visibility × next-item click × return
Once those stages are separated, “foreign traffic = 0” stops being mysterious.
Because a zero that has a location can be repaired.
References
https://developers.google.com/search/docs/fundamentals/how-search-works
https://developers.google.com/search/docs/specialty/international/localized-versions
https://developers.google.com/search/docs/crawling-indexing/links-crawlable
References (3)
- Google Search Central, In-depth guide to how Google Search works
- Google Search Central, Localized versions of your pages
- Google Search Central, Link best practices for Google
