Direct answer: The 36 article pages contain 445 body paragraphs (20 characters or more), a median of 12 per page and a median length of 58 characters. Eleven pages contain 31 paragraphs longer than 300 characters, the longest 982 characters, and none of the 36 pages uses a <table>. Median visible text is 1,919 characters.
Another 22 of 36 pages have no <h2>, matching the study-53 heading-outline pilot; this study does not re-analyse headings and only treats them as context. E19 has 1,610 characters of text but no <p> of 20 characters or more, which suggests the copy is laid out in divs.
Method: reuse the study-45 article cache
The study parses the 36 article pages cached on 2026-11-03 08:40–08:41 (10 sites, HTTP 200, HTML ≥5 KB) with zero network requests. Only <p> elements with 20 or more visible characters count as paragraphs; scripts, styles and tags are removed, entities decoded and whitespace collapsed. A long paragraph is over 300 characters.
| Item | Rule | Boundary |
|---|---|---|
| Paragraph | p with 20+ visible characters | Div/br layout not counted |
| Long paragraph | Over 300 characters | Readability not judged |
| Lists and tables | Counts of ul/ol/li/table/tr | Navigation and footer included |
| Heading density | (h2+h3) ÷ visible characters × 1,000 | Quality not judged |
| Text length | Visible characters after tag removal | Includes navigation and footer text |
Result: few, short paragraphs
| Measure | Result |
|---|---|
| Paragraphs | 445 |
| Median per page | 12 |
| Median length | 58 characters |
| Longest paragraph | 982 characters |
| Paragraphs over 300 characters | 31 on 11 pages |
| Median visible text | 1,919 characters |
| Median heading density | 0.1 per 1,000 characters |
| Tables | 0 |
| List items | 2,197 in 596 list containers, mostly navigation |
List counts cannot stand for body structure: most of the 596 ul/ol containers are navigation menus, and page-level counting does not separate body from template. This study therefore treats only tables and paragraphs as structural evidence and records lists for completeness.
Long paragraphs on 11 pages
| Site | Page | Long paragraphs | Longest (characters) |
|---|---|---|---|
| E03 | 1–3 | 各 3 | 400 |
| E03 | 4 | 3 | 982 |
| E09 | 2 | 2 | 312 |
| E14 | 3 | 1 | 416 |
| E14 | 4 | 10 | 400 |
| E22 | 1 | 3 | 326 |
| E22 | 2 | 1 | 313 |
| E22 | 3 | 1 | 314 |
| E23 | 2 | 1 | 307 |
E14's fourth page carries 10 long paragraphs (17 paragraphs, 4,923 characters), the densest in the sample; E03's four pages each carry three, including one of 982 characters. E22 has one to three per page. A long paragraph is not automatically wrong, but it blurs boundaries for mobile scanning and AI extraction.
Tables: zero across 36 pages
None of the 36 article pages uses a <table>. That differs from homepages, where tables were one of the main text carriers in the study-58 text-share pilot. For equipment parameters, models and operating conditions, readers must dig parameters out of prose.
Relationship to headings
22 of 36 pages have no <h2> (recorded in study 53), and median heading density here is 0.1 per 1,000 characters. Paragraph structure therefore reads like a continuous article: few paragraphs, short ones, almost no subheads. E19 is the opposite extreme — no qualifying <p> at all, with text likely built from divs or line breaks.
What this means for acceptance
- Put paragraph length into acceptance. A median under 150 characters and no paragraph over 300 is a reasonable bar; 31 paragraphs here exceed 300.
- Use tables for parameters. Zero tables across 36 pages; models, conditions and capacity are easier to read and extract in a table.
- Break long copy with subheads. 22 pages have no H2 and heading density is 0.1 per 1,000 characters.
- Check whether body copy uses
p. E19 has 1,610 characters but no qualifying paragraph and needs a manual look. - Do not count navigation lists as text structure. Most of the 596 list containers come from the template.
Limits
- Paragraph rule: only
<p>with 20+ characters counts; div/br layouts are undercounted. - Page level: navigation and footer text enter text length and list counts; body is not separated.
- No readability judgement: paragraph length is a proxy, not a comprehension measure.
- One snapshot: cached on 2026-11-03; redesigns are not rechecked.
- 10 sites, 36 pages: a pilot, no industry claim.
Data and reproduction
Anonymised data: equipment-article-body-structure-2026.csv (36 rows × 15 columns: text length, paragraphs, median and longest length, long counts, lists and items, tables and rows, H2/H3, heading and paragraph density). The repository keeps docs/research/2026-11-article-body-structure.csv.
Parser: scripts/article_structure_probe.py; unit tests: scripts/test_article_structure_probe.py (no network, eight tests). Run:
python -m unittest scripts.test_article_structure_probe
Related: the article heading-outline pilot, the page text-share pilot, and the article anchor-text pilot., and the article outbound-link pilot, and the article body-link pilot, and the body link checker
Sources and definitions
- Article cache: 36 pages from 10 sites in study 45 (2026-11-03 08:40–08:41).
- Paragraph:
pwith 20+ visible characters; long paragraph over 300 characters. - Scan: 2026-11-06, zero network requests.
Related: the article HTML weight composition pilot.
Related: the article visible-text within-site repetition pilot.