Direct answer: The 25 cached homepages contain 2,739 <a> links: 2,326 internal, 240 external, 13 mailto:, 43 tel:, 40 in-page anchors and 77 javascript:. Of these, 344 use target="_blank", 100 carry rel="nofollow" and 391 have no visible text. Another 3 homepages contain no links at all (E18, E20, E25).
Earlier studies covered images and structure; this one looks at links: internal and external, special protocols, new windows, nofollow and links without visible text. Still zero requests on cached HTML.
Method: classify by href
Using the 25 cached homepages from study 4, every <a> with href is parsed: same-host absolute URLs and relative paths count as internal, other hosts as external, with mailto, tel, in-page anchors and javascript counted separately. The study also counts target="_blank", rel containing nofollow, and links whose inner HTML has no visible text.
| Item | Rule | Boundary |
|---|---|---|
| Internal / external | Compared with the homepage host | Subdomains judged by string containment |
| mailto / tel | Protocol prefix | Number validity not checked |
| No visible text | Inner tags removed, result empty | An inner image alt may still name it |
| nofollow | rel contains nofollow | Sponsored/ugc not separated |
Result: internal links dominate, 240 external
| Type | Count | Sites (25) |
|---|---|---|
| Internal | 2,326 | 22 |
| External | 240 | 21 |
mailto: | 13 | 7 |
tel: | 43 | 11 |
| In-page anchor | 40 | 7 |
javascript: | 77 | 15 |
Internal links are 84.9% and external 8.8%. The 77 javascript: links show some interactions still use script placeholders; they are unfriendly to crawlers and keyboard users, but this study only counts them.
New windows and nofollow
target="_blank": 344, across 22 sites; E27 has 82 on one page and E03 has 55.rel="nofollow": 100, across 9 sites; E26 has 64, E03 11 and E14 7.
The study did not count rel="noopener", so it cannot judge new-window security; it only shows that new windows are common.
391 links have no visible text
| Site | Links without visible text | Note |
|---|---|---|
| E13 | 56 | Mostly icon or image links |
| E23 | 44 | — |
| E17 | 38 | — |
| E06 / E26 | 34 each | — |
| Total | 391 | Across 22 sites |
No visible text does not mean no accessible name: if the link contains an image with alt, a screen reader can still announce it. Deciding whether the link is truly unnamed requires combining alt and aria-label; study 44 does that merge and finds 254 of the 391 named and 137 unnamed.
Three homepages have no links
E18, E20 and E25 contain no <a href> in the cached HTML. They may be display-only pages, use JavaScript routing, or show only images and scripts. Either way, static HTML exposes no crawlable link entry, which acceptance should confirm separately.
What this means for acceptance
- Record internal and external separately. There are 240 external links across 21 sites.
- Combine no-visible-text with alt/aria. The 391 are “no text”, not a final verdict.
- New windows are common. 344 use target=_blank; confirm noopener and related attributes.
- nofollow is not an error. The 100 sit on a few sites and may be intentional.
- List the linkless homepages. Three sites need confirmation that navigation is not JavaScript-only.
Limits
- Cached, zero-request: pages are frozen at 2026-10-31 to 11-01.
- Regex parsing: nested anchors, commented links and JavaScript-inserted links are not handled.
- Host check: string containment may treat lookalike domains as internal.
- No accessible-name check: a link without text may still be named by an image alt.
- 25-site pilot: no industry claim.
Data and reproduction
Anonymised data: equipment-link-structure-2026.csv (25 rows: links, internal, external, mailto, tel, anchors, javascript, target=_blank, nofollow, no-visible-text). The repository keeps docs/research/2026-11-link-structure.csv.
Parser: scripts/link_structure_probe.py; unit tests: scripts/test_link_structure_probe.py (9 tests, no network). Run:
python -m unittest scripts.test_link_structure_probe
Related: the homepage semantics pilot, the homepage image alt coverage and text study, the homepage image attributes pilot, and the homepage link accessible names pilot, and the article page link structure pilot.
Sources and definitions
- Homepage cache: 25 homepages from study 4 (2026-10-31 to 11-01).
- Link rules: href classification, target/rel attributes, visible inner text.
- Scan: 2026-11-03, zero network requests.