Google Traffic Drops After Migrating an English Site to a Subdomain: A Complete Audit From 2,923 Unindexed Pages to GLM-5.2 Retranslation Samples

In July 2026, I migrated the English version of the blog from the original:

https://www.shuijingwanwq.com/en

to an independent subdomain:

https://en.shuijingwanwq.com

The old /en/ URLs were uniformly redirected to the new English subdomain via 301.

After the migration, the English site itself remained accessible, and issues related to Cloudflare, Polylang, canonical, and caching were resolved one after another. However, on August 8, 2026, when I checked Google Search Console again, I noticed a problem I had not fully paid attention to before:

The English site had almost no clicks from Google search.

Further investigation revealed that the problem was not that Google couldn’t find the English pages, nor that the Sitemap hadn’t been submitted, let alone that robots.txt or noindex was blocking indexing on a large scale.

What truly deserves attention is:

Google had already discovered and crawled a large number of English pages, but the vast majority still had not entered the index.

Moreover, even pages that had already been retranslated by GLM-5.2 and recrawled by Google were still generally in the “Crawled – currently not indexed” state.

This article documents the entire troubleshooting process.

1. The English site had almost no Google clicks in the last 7 days

First, I checked the English site:

https://en.shuijingwanwq.com

in the “Insights” section of Google Search Console.

The past 7 days directly showed:

No clicks during this period.

Meanwhile, the “Page indexing” page showed:

• Not indexed: 7,921
• Indexed: 273

This means that among the English site URLs currently known to Google, only a very small fraction actually entered the index.

At first, I suspected:

This might be the direct reason why the English site currently has almost no organic Google search traffic.

Because if a page has not entered the Google index at all, subsequent keyword ranking, impressions, and clicks are naturally out of the question.

However, the number 7,921 itself cannot directly represent “7,921 English articles are not indexed.”

WordPress itself generates a large number of:

• tag pages;
• category pages;
• various archive pages;
• redirect URLs;
• parameterized URLs;
• pages actively set to noindex.

Therefore, the next step was to look at why Google does not index these pages.

2. Among the 7,921 unindexed pages, the real concern is “Crawled – currently not indexed”

Google Search Console provided 9 categories of reasons for non-indexing:

• Excluded by noindex: 4,488
• Alternate page with proper canonical tag: 27
• Page with redirect: 14
• Not found (404): 13
• Server error (5xx): 2
• Duplicate without user-selected canonical: 1
• Crawled – currently not indexed: 2,931
• Discovered – currently not indexed: 439
• Duplicate, Google chose different canonical than user: 6

Although the 4,488 noindex pages make up the largest number, they are not what I am most worried about.

Because the blog inherently has many tag and archive pages, it is normal for a lot of them to actively block search engine indexing.

What truly deserves attention is:

Crawled – currently not indexed: 2,931

This means Google not only knows the URLs but has actually visited the pages; it just didn’t add them to the formal search index after crawling.

In contrast, the 439 “Discovered – currently not indexed” pages might at least just not have reached the crawling stage yet.

So the subsequent investigation focused on these 2,931 pages.

3. The 2,931 pages do indeed include a large number of English article bodies

After entering “Crawled – currently not indexed”, I could see a large number of standard article URLs, such as:

https://en.shuijingwanwq.com/2022/07/16/13352

https://en.shuijingwanwq.com/2015/08/24/16359

https://en.shuijingwanwq.com/2020/01/10/15022

https://en.shuijingwanwq.com/2026/06/24/18173

Of course, the list also mixed in:

• tag pages;
• category pages;
• other archive pages.

But it could already be confirmed:

The 2,931 “Crawled – currently not indexed” pages are not just auxiliary pages automatically generated by WordPress; they do indeed include a large number of English article bodies.

4. Spot-checking articles: crawling, indexing permissions, and canonical are all normal

Next, I randomly checked a few English articles from different years.

First:

https://en.shuijingwanwq.com/2022/07/16/13352

Search Console showed:

• Status: Crawled – currently not indexed
• Last crawl: August 8, 2026
• Googlebot: Smartphone
• Crawl allowed?: Yes
• Page fetch: Successful
• Indexing allowed?: Yes
• User-declared canonical: Current English URL
• Google-selected canonical: Inspected URL

Then I checked:

https://en.shuijingwanwq.com/2015/08/24/16359

The results were almost exactly the same:

• Crawl successful;
• Indexing allowed;
• canonical points to itself;
• Google also recognizes the current URL as canonical;
• but it is still “Crawled – currently not indexed”.

I then checked a newer article:

https://en.shuijingwanwq.com/2026/06/24/18173

The result was still the same.

At this point, I could at least lower the priority of several technical issues:

• robots.txt blocking crawling on a large scale;
• pages setting noindex;
• page crawl failures;
• canonicals being universally wrong;
• Google selecting the Chinese original as canonical.

None of these phenomena appeared in the spot-checked pages.

5. The old /en/ actually used to have Google search traffic

This raised another question:

When the English site was previously placed under the /en/ directory, did it actually have any Google search performance?

So I switched to:

https://www.shuijingwanwq.com

the corresponding Search Console property, and set the following in “Performance”:

• Time: Last 3 months;
• Page filter: URL contains /en/.

The results showed:

• Total clicks: 134
• Total impressions: 71,800
• Average CTR: 0.2%
• Average position: 14

This indicates a very important point:

Before migrating to the independent subdomain, the English pages were not entirely absent from Google search.

Although 134 clicks is not much, and a 0.2% CTR is relatively low, 71,800 impressions are enough to prove that:

Google had indeed indexed and displayed a considerable number of /en/ English pages before.

Moreover, after the migration occurred around July 10, the impressions and clicks for the old /en/ did not immediately drop to zero.

This means that a portion of the Google traffic currently accessing the new English site is likely still going through:

Google 搜索结果中的旧 /en/ URL

↓

301

↓

en.shuijingwanwq.com

That is to say:

Users ultimately access the new English subdomain, but some of Google’s historical search signals still come from the old /en/ URLs.

6. Checking the old URL: the 301 itself works normally

Using this article as an example:

Old address:

https://www.shuijingwanwq.com/en/2026/06/24/18173

New address:

https://en.shuijingwanwq.com/2026/06/24/18173

After checking the old URL, Google Search Console clearly showed:

Page not indexed: Page with redirect

And Google already knew its canonical target was:

https://en.shuijingwanwq.com/2026/06/24/18173

This shows that the migration link itself is normal, at least for this sample:

旧 /en/ URL

↓

301

↓

新 en. URL

The real problem is not that the old pages are still holding index spots, but rather:

The old pages have exited, but the new pages have not fully taken over yet.

7. The Sitemap has also been properly submitted and read by Google

Next, I checked the English site’s Sitemap.

Search Console showed:

• Sitemap: /sitemap_index.xml
• Submitted date: July 12, 2026
• Last read: August 7, 2026
• Status: Success
• Discovered URLs: 3,210

Therefore, it is basically possible to rule out:

Google doesn’t know the English articles exist because there is no Sitemap.

Google has already discovered 3,210 pages from the Sitemap.

8. Looking only at submitted Sitemap pages, the indexing rate is still only about 8%

To eliminate the interference from a large number of noindex tag pages and other non-target URLs, I switched “Page indexing” from:

All known pages

to:

All submitted pages

The results became more intuitive:

• Not indexed: 2,923
• Indexed: 252

That is, among the 3,175 submitted pages in the current report:

only about 8% have entered the index.

Looking again at the reasons for the 2,923 unindexed pages:

• Excluded by noindex: 19
• Server error (5xx): 1
• 404: 1
• Alternate page: 1
• Crawled – currently not indexed: 2,456
• Discovered – currently not indexed: 439
• Google chose different canonical: 6

Among them:

2456 + 439 = 2895

This means that almost all unindexed submitted pages are concentrated in:

Pages that Google has already discovered or crawled, but has not yet officially brought into the index.

Issues genuinely caused by:

• noindex
• 5xx
• 404
• canonical

are only a very small minority.

At this point, it is quite difficult to explain the problem as a traditional technical SEO configuration error.

9. A page that previously had 10,659 impressions also did not re-enter the index after migration

To further determine whether Google was simply eliminating pages that originally had no search value, I found one of the best-performing pages in the Search Console data for the old /en/:

https://www.shuijingwanwq.com/en/2026/03/02/9744

Over the past 3 months:

• Clicks: 15
• Impressions: 10,659

The corresponding new URL:

https://en.shuijingwanwq.com/2026/03/02/9744

The inspection result was still:

Crawled – currently not indexed

Moreover, Google had just recrawled the page on August 8, 2026:

• Crawl successful;
• Indexing allowed;
• canonical correct;
• Google also selected the current new URL.

This shows that:

The problem cannot simply be explained as “Google just no longer needs those originally worthless old articles”.

Because at least this article previously had very clear search exposure.

10. Live URL test is completely normal

For the above article:

https://en.shuijingwanwq.com/2026/03/02/9744

I also performed Search Console’s “Test Live URL”.

The result showed a green status:

URL can be indexed by Google

At the same time:

• Page availability is normal;
• Page can be indexed;
• Enhancements can also be detected normally.

This further indicates that:

There is currently no obvious real-time technical indexing obstacle for the page.

The historical status in Search Console being “Crawled – currently not indexed” does not mean the current page cannot be indexed due to a technical error.

11. Is the old AutoPoly / Chrome AI translation quality the cause?

Investigating to this point, there is a very natural suspicion:

Many of my old English articles were originally generated via AutoPoly and similar methods, most of which once used Chrome’s built-in AI translation.

The quality of these old translations was uneven.

Now the blog is undergoing retranslation of historical English articles, using GLM-5.2 to regenerate from the Chinese original:

• English titles;
• English summaries;
• English body text.

Therefore, it is necessary to further confirm:

Is what Google sees now still the old Chrome AI translation, or the updated GLM-5.2 translation?

If Google has not recrawled yet, then the old translation quality could still directly explain the current state.

If Google has already recrawled after the GLM overwrite, then the situation is different.

12. Randomly selecting 3 articles that completed GLM overwriting on July 22

From the historical migration repository:

wordpress-ai-excerpt-backfill

I found the batch:

syntaxhighlighter-20260722-01

and randomly selected 3 articles:

9376
15485
15961

9376
15485
15961

The corresponding status records showed:

First article

Chinese article:

9376

English article:

11220

Status:

workflow_status: completed

Second article

Chinese article:

15485

English article:

15551

Status:

workflow_status: completed

Third article

Chinese article:

15961

English article:

16048

Status:

workflow_status: completed

The corresponding English URLs are:

https://en.shuijingwanwq.com/2026/03/25/11220

https://en.shuijingwanwq.com/2026/06/02/15551

https://en.shuijingwanwq.com/2026/06/06/16048

These articles completed their historical overwrite translation on July 22.

Therefore, as long as Google’s last crawl time is later than July 22, what it sees is no longer the original AutoPoly / Chrome AI translation, but the new content after the GLM-5.2 overwrite.

13. All 3 GLM-5.2 retranslated articles still have not entered the index

I checked the 3 URLs separately.

First article:

https://en.shuijingwanwq.com/2026/03/25/11220

Google last crawl:

July 30, 2026

Status:

Crawled – currently not indexed

Second article:

https://en.shuijingwanwq.com/2026/06/02/15551

Google last crawl:

August 8, 2026

Status:

Crawled – currently not indexed

Third article:

https://en.shuijingwanwq.com/2026/06/06/16048

Google last crawl:

August 3, 2026

Status:

Crawled – currently not indexed

All 3 articles met the following conditions:

• Completed GLM-5.2 overwrite translation;
• Google crawl time later than July 22;
• Crawl successful;
• Indexing allowed;
• canonical normal;
• Google selected the current URL as canonical;
• Still did not enter the index.

Therefore, “poor old AutoPoly / Chrome AI translation quality” can no longer serve as the main direct explanation for the current large-scale unindexed phenomenon.

The old translation quality might have previously affected the content quality signals the English site accumulated over time, but Google has now reviewed the GLM-5.2 version, and the pages still have not immediately restored their index status.

14. This currently looks more like re-indexing and re-evaluation after the /en/ → en. migration

After this round of troubleshooting, the current chain of evidence is roughly as follows:

Old English site /en/:

• Previously had real Google search impressions and clicks;
• Had 71,800 impressions in the past 3 months;
• 134 clicks;
• Some individual pages had tens of thousands of impressions.

After migration:

• Old /en/ URLs correctly 301;
• Google has recognized the redirect relationship;
• New en. URLs are accessible normally;
• robots is not blocking;
• Pages do not have noindex;
• Crawl successful;
• canonical normal;
• Google’s selected canonical is also the current English URL;
• Sitemap submitted and successfully read;
• Live URL test confirms it can be indexed.

But at the same time:

• Only 252 submitted pages have entered the index;
• 2,456 are “Crawled – currently not indexed”;
• 439 are “Discovered – currently not indexed”.

Even:

• The page that previously had 10,659 search impressions also did not smoothly take over the index;
• 3 randomly selected articles retranslated by GLM-5.2 on July 22 also all failed to enter the index after being recrawled.

Therefore, the explanation that best fits the current evidence is:

Google is recrawling the entire English site migrated from /en/ to en., re-establishing index relationships, and re-evaluating site-level and page-level quality, and the index takeover speed is clearly lagging behind the crawl speed.

There might also be a transitional “index gap” here:

The old URLs have gradually exited the index due to 301,

but the corresponding new URLs have not immediately completed the index takeover.

15. Temporarily stopping the search for non-existent “technical faults”

Having troubleshot to this point, I decided to temporarily stop randomly checking more URLs.

Because the different types of samples currently available are highly consistent:

• 2015 historical articles;
• 2022 historical articles;
• 2026 new articles;
• important articles that previously received tens of thousands of impressions;
• articles already retranslated by GLM-5.2 and recrawled.

All showed similar results.

Continuing to spot-check the 4th, 10th, or 20th article would likely just repeat the same conclusion.

There is also no need right now to temporarily disrupt the ongoing historical article retranslation plan just for the sake of Search Console data.

The historical English retranslation continues:

AutoPoly / Chrome AI old translations
→ GLM-5.2 high-quality retranslation

This work itself is continuously improving the English site’s content quality.

16. The index trend is more worth observing next

In the next phase, I plan to focus on observing a few metrics rather than continuing to frequently modify the site configuration.

First, the number of “Indexed” pages on the English site.

Currently it is:

252

If it gradually changes to:

400 → 700 → 1000 → 1500……

then it basically proves that the current situation is mainly just a slow index takeover speed after the migration.

Second:

Crawled – currently not indexed: 2,456

Observe whether this number gradually decreases.

Third:

Discovered – currently not indexed: 439

Observe whether Google continues to complete the crawling.

Fourth, whether the Search Console for the new English subdomain starts showing:

• impressions;
• clicks;
• search queries;
• rankings.

Fifth, whether the search impressions for the old /en/ gradually decrease, while the corresponding signals gradually transfer to:

en.shuijingwanwq.com

Summary

This troubleshooting initially started just because I noticed:

The English site’s Search Console surprisingly had no Google clicks in the past 7 days.

But upon digging deeper, I found that the real problem is much more fundamental than “low traffic”.

The biggest current issue for the English site is:

Google has already discovered and even crawled a large number of pages, but they still have not entered the formal index.

Initially, I suspected:

• robots.txt;
• noindex;
• Sitemap;
• canonical;
• 301;
• Cloudflare;
• page crawl anomalies;
• old AutoPoly / Chrome AI translation quality.

After checking each item, no obvious technical fault was found that could explain the 2,456 “Crawled – currently not indexed” pages.

In particular, after randomly selecting 3 historical articles that had already been overwritten by GLM-5.2 translation on July 22, Google recrawled these pages on July 30, August 3, and August 8 respectively, but they all still remained “Crawled – currently not indexed”.

Therefore, the more reasonable interim conclusion right now is:

After the English site migrated from /en/ to the independent en. subdomain, Google’s re-indexing and search signal takeover have clearly lagged behind the page crawling progress.

The old AutoPoly / Chrome AI translation quality might still be part of the historical content quality background of the English site, but it can no longer solely explain why these pages are still not indexed.

Next, rather than continuing to constantly modify SEO configurations, it is more worthwhile to:

Continue advancing the existing GLM-5.2 historical English retranslation, while observing whether the number of indexed pages in Search Console starts to clearly recover over the next few weeks.

If the number of indexed pages starts to grow continuously, it means the whole migration just needs time.

If Google continues to recrawl these already high-quality retranslated pages over the long term, and the indexing ratio still stays at the current level of about 8%, then it will be necessary to further investigate in the next phase:

• the overall content quality signals of the English site;
• internal link structure;
• the degree of duplication between pages;
• the independent search value of the English pages;
• Google’s site-level quality assessment of the entire English subdomain.

At least as of August 8, 2026, this troubleshooting has basically confirmed:

The English site is not “invisible to Google”, but rather “Google has seen it, has crawled it, but a large number of pages temporarily have not re-entered the index”.

This also explains why the English site, despite clearly having thousands of articles, has almost no Google search traffic from the new en. subdomain itself.