English-Site Recheck: Path Expansion Finds 9 of 25 English Sites, Five New

Direct answer: with only /en, four of 25 sites are genuinely English; after adding six entries (including /en-us and the en. subdomain), five new English sites appear (E03, E08, E09, E11, E22) for 9 of 25. Of the eight /en pages, two are Chinese fallbacks, one a WAF block and one a 404; none of the four English pages declares hreflang.

This study discovers English entry points from homepage links, hrefs and hreflang, then requests once; it does not run JavaScript, click menus or judge translation quality.

Method: link discovery plus one /en fallback

On 2026-11-08 between 13:40 and 13:42 each of the 25 homepages from study 4 received one GET; candidates came from <a> text (EN/English/英文), /en in the href and hreflang="en*", with at most three tried. If none succeeded, /en/ and /en were tried. For a successful candidate we read html lang, whether the title contains Chinese and whether hreflang is declared. Pages are not rendered.

Result: two switches, eight /en pages, nine English sites after expansion

Of the eight reachable /en pages, only two expose a homepage switch; 200 does not mean usable English: E18 is a WAF shell, E27 a 404, and E04 and E12 return Chinese bodies. Extending discovery to subdomains and common paths lifts English sites from four to nine — testing only /en undercounts by more than half.

Content check: four English, two Chinese, one WAF, one 404

On 2026-11-08 between 14:41 and 14:44 we read the visible text of each of the eight /en pages, counted CJK characters and English words, and detected WAF and 404 pages before the language verdict.

The verdict uses visible text: zero CJK characters counts as English; up to 50 CJK characters with at least 50 English words counts as mostly English; anything with substantial Chinese counts as Chinese. WAF and 404 pages are identified before the language verdict so a “200 error page” is not counted as an English site.

Path expansion: six paths add five English sites

On 2026-11-08 between 15:40 and 15:43 each of the 25 homepages was tried with six entries: /en-us/, /en/index.html, /english/, /global/, /en/home and the en.host subdomain (8–12 s timeout). Pages returning 200 were classified by visible text before counting.

All five new English sites sit outside /en: E03, E08, E09 and E22 use the en. subdomain, and E11 serves both /english/ and the subdomain. Testing only /en undercounts the English presence by more than half (4 to 9).

Language signals: four English pages, zero hreflang

lang and hreflang are separate signals: E01 has an English title but lang=zh-CN; E29 has an English title but no lang. Zero hreflang means the Chinese and English pages do not declare each other, leaving search engines to guess from URLs and links.

What this means for acceptance

1. Put a text language switch on the homepage. Only two of the eight sites with /en expose one; an EN/English text link is more discoverable than a flag icon.
2. /en must return English, not a soft fallback or error page. E04 and E12 return Chinese bodies, E18 is a WAF shell and E27 a 404; status 200 is not proof of usable content.
3. Set lang=en on English pages. E01's zh-CN and the missing values on E27/E29 mislead screen readers and search engines.
4. Declare hreflang both ways. None of the four /en English pages does, consistent with the homepage canonical and scheme-redirect pilot hreflang coverage.
5. Cover subdomains and common paths when discovering entries. /en alone found four English sites; six paths found nine, mainly through the en subdomain and /english.
6. Include the English entry in export-site acceptance. For equipment exporters, the entry, hreflang and lang belong in one checklist; see the website development service.

Limits

• Text and URL only: image buttons, JS menus and unlabelled flag icons are not counted, so switches may be underestimated.
• Content quality not judged: an English title does not prove a full translation; text, product data and contact details were not checked.
• Path coverage is finite: five common paths and the en subdomain were added, but other subdomains and names remain untested; the five new sites were only language-classified, not checked for hreflang/lang.
• One snapshot: 2026-11-08 13:40–13:42; E19's homepage timed out this run while its /en was reachable.
• 25-site pilot: no industry claim.

Data and reproduction

Anonymised data: equipment-homepage-english-entry-2026.csv (25 rows × 11 columns: switch, /en status, lang, title, hreflang). Content check: equipment-homepage-english-content-2026.csv (8 rows × 10 columns: CJK characters, English words, verdict). Path expansion: equipment-homepage-english-paths-2026.csv (25 rows × 8 columns: paths tried, 200s, new English, path type). The repository keeps the matching CSVs under docs/research/.

Parser: scripts/language_entry_probe.py; unit tests: scripts/test_content_language_entry.py (9 tests, no network). Run:

python -m unittest scripts.test_content_language_entry

Related: the homepage canonical and scheme-redirect pilot, the homepage entity and compliance signal pilot, and the homepage enquiry-path pilot.

Sources and definitions

1. Sample: the 25 homepages from study 4 (same sample_id).
2. Fetch: 2026-11-08 13:40–13:42 Asia/Shanghai, serial, 0.5 s apart, 15/12-second timeouts.
3. Candidates: link text, /en in the href, hreflang=en*; fallback only /en/ and /en.
4. Metrics: HTTP status, html lang, whether the title contains Chinese, whether the English page declares hreflang.
5. Content check: 2026-11-08 14:41–14:44; visible-text CJK characters and English words, with WAF/404 detected before the language verdict.
6. Path expansion: 2026-11-08 15:40–15:43; six entries per site (five paths plus the en subdomain), with 200 pages classified by visible text.

Frequently asked questions

Why do only two sites show a switch?

This study recognises link text (EN/English/英文), /en in the href and hreflang only; image buttons, JavaScript menus and unlabelled flag icons are not counted, so the count may be low. The five sites found by path expansion also show that “no homepage entry” does not mean “no English site”.

Does a 200 on /en mean an English site?

No. E12 and E18 return 200 on /en but keep a Chinese title, a soft fallback; only E01, E19, E24, E27 and E29 have English titles.

What does zero hreflang affect?

Search engines struggle to pair the Chinese and English pages as language versions of one entity; multilingual sites should declare hreflang both ways.

Is lang=zh-CN on an English page a problem?

It is a conflicting signal: screen readers and search engines treat the page as zh-CN and may misidentify the English content; use en or en-US.

Do 25 sites represent the industry?

No. This is the same 25 equipment-company homepages from study 4; it describes only these pages.

Can the data be re-checked?

Yes. The entry CSV has 25 rows by 11 columns, the content check 8 by 10 and the path expansion 25 by 8; the two modules have 20 offline unit tests.

How are WAF and 404 pages detected?

E18's /en returns 200 but its visible text is a WAF script shell; E27's /en leads to a page titled 404. Both are treated as unusable and not counted as English sites.

Why try /en-us and subdomains?

Many sites host English on an en subdomain or /en-us, /english paths; testing only /en undercounts. The six-path pass added five sites.