Direct answer: Across 10 sites and 36 article pages, 8 pages have no H1 (22.2%) and 4 have three or four H1s. Among the 28 pages with an H1, 20 have an H1 contained in the title, while 8 H1s are unrelated to the title: five are the placeholder “这里是标题一h1占位文字” and three say “联系我们” on unrelated articles.
Other head fields: all 36 pages have a title, median 23 characters. All have a meta description, but 18 are under 50 characters (shortest 4) with a median of 60. Six pages lack canonical (the 30 present are all self-referential), five lack lang, and six lack OG title and description. No page declares a robots meta tag.
Method: reuse the raw HTML from study 45
The sample matches study 45: the article frame from study 28, 10 sites and 36 usable article pages (five JavaScript shells from another site are listed separately). This study does not refetch; it reads the raw HTML saved on 2026-11-03 between 08:40 and 08:41, with zero network requests.
Regex extracts title, every H1, meta description (both attribute orders), canonical, html lang, og:title, og:description and meta robots. The H1/title relation has four classes: exact, contained, unrelated and missing; containment only means shared text, not shared meaning.
| Item | Definition | Boundary |
|---|---|---|
| H1 count | All <h1> elements on the page | Visual size and position not checked |
| H1/title relation | Exact / contained / unrelated | Containment is not semantic agreement |
| Description length | Characters after tags, entities and whitespace | Length is not quality |
| Canonical self-reference | Host without www and path without trailing slash | Redirects not followed |
Result: H1 is the biggest issue
| H1 count | Pages | Note |
|---|---|---|
| 0 | 8 | 22.2% |
| 1 | 24 | 66.7% |
| 3–4 | 4 | One site, up to 4 on a page |
| H1 vs title | Pages | Note |
|---|---|---|
| Exact | 0 | Titles add a brand suffix; equality is not required |
| Contained | 20 | H1 is part of the title, or vice versa |
| Unrelated | 8 | 5 placeholders, 3 “联系我们” |
| No H1 | 8 | — |
The placeholders sit on five E09 pages; three E23 pages use “联系我们” as the H1 for history, industry news and news listing pages. These are not “could be better” issues: the template either was not replaced or reuses the wrong block, and acceptance should fail them.
Descriptions exist, but half are short
| Description length | Pages | Note |
|---|---|---|
| < 50 chars | 18 | 50.0%; shortest 4 |
| 50–99 chars | 9 | — |
| 100–136 chars | 9 | Longest 136 |
| Total | 36 | Median 60 |
A four-character description is effectively none; five 15-character ones come from one site. Length is not the only criterion, but a very short description leaves the search engine to assemble its own snippet.
Canonical, lang and OG
| Item | Present | Missing | Note |
|---|---|---|---|
| canonical | 30 | 6 | All 30 are self-referential |
| html lang | 31 | 5 | All present values are zh |
| og:title | 30 | 6 | Same pages as canonical |
| og:description | 30 | 6 | Same pages |
| meta robots | 0 | 36 | Not declared; default index assumed |
By site
| Site | Pages | No H1 | Multi H1 | H1 issue | No canonical / lang | Short desc |
|---|---|---|---|---|---|---|
| E03 | 5 | 1 | 4 | 0 | 0 / 0 | 5 |
| E06 | 5 | 0 | 0 | 0 | 0 / 0 | 0 |
| E09 | 5 | 0 | 0 | 5 placeholder | 0 / 0 | 0 |
| E14 | 5 | 0 | 0 | 0 | 0 / 0 | 5 |
| E16 | 3 | 3 | 0 | 0 | 3 / 3 | 0 |
| E19 | 1 | 0 | 0 | 0 | 1 / 1 | 1 |
| E22 | 5 | 1 | 0 | 0 | 0 / 0 | 1 |
| E23 | 5 | 2 | 0 | 3 wrong | 0 / 0 | 5 |
| E26 | 1 | 0 | 0 | 0 | 1 / 0 | 0 |
| E28 | 1 | 1 | 0 | 0 | 1 / 1 | 1 |
Only E06's five pages have no issue in H1, canonical, lang or description length. That does not make the pages fully compliant; it only means these fields are clean.
What this means for acceptance
- Fail placeholder and wrong H1s. Five placeholders and three unrelated H1s come from unreplaced template blocks.
- Add the missing H1s first. 22.2% of article pages have no single main heading.
- Reduce multiple H1s. Four pages have three or four, usually because sidebar headings reuse H1.
- Do not accept any description. 18 pages are under 50 characters, four have only 4.
- Check canonical, lang and OG together. The missing pages overlap, so one template fix covers them.
Limits
- Cached, zero-request: the HTML was fetched on 2026-11-03 08:40–08:41; later changes are not reflected.
- Regex parsing: comments, script text and dynamically inserted head tags are not handled.
- Coarse relation: “contained” only means shared text, not shared meaning; H1 position and visual level are not measured.
- Length is not quality: character count says nothing about readability or keyword coverage.
- 10 sites, 36 pages: a pilot, no industry claim.
Data and reproduction
Anonymised data: equipment-article-meta-2026.csv (36 rows × 12 columns: title length, H1 count and relation, description length, canonical, lang, OG and robots). The repository keeps docs/research/2026-11-article-meta.csv. Raw HTML stays off-site and is not deployed.
Parser: scripts/article_meta_probe.py; unit tests: scripts/test_article_meta_probe.py (7 tests, no network). Run:
python -m unittest scripts.test_article_meta_probe
Related: the article page link structure pilot, the article visible date pilot, and the homepage anchor text pilot.
Sources and definitions
- Sample: the study 28 article frame, 36 usable pages fetched by study 45 on 2026-11-03 08:40–08:41.
- Head fields: title, every H1, description, canonical, lang, OG and robots.
- H1 relation: exact / contained / unrelated / missing, text only.
- Scan: 2026-11-03, zero network requests.