Direct answer: of 25 equipment-company homepages, only 9 declare rel=canonical; the other 16 do not. All 9 canonical values exactly match the final URL (same scheme, host, path and trailing slash). Only 8 have og:url, and 8/8 match the canonical; E17 has a canonical but no og:url. Only E24 declares hreflang, two entries.
Missing canonical does not mean a page cannot be indexed, but it leaves the consolidation of www/non-www, http/https and parameter variants to the search engine. Four homepages were served over http:// (E04, E13, E16, E27), and all four also lack canonical; those sites have more protocol/host duplication risk.
Method: cached HTML only, exact string comparison
We reused the 25 homepage caches from study 4 (2026-10-31) and parsed only static <link rel="canonical"> and <meta property="og:url"> from the HTML. The comparison target is the final URL recorded at capture time (after redirects). Comparison is an exact string match: scheme, www, case, trailing slash, and query parameters are not normalised. No rendering, no JavaScript execution and no network requests.
| Item | Rule | Boundary |
|---|---|---|
| Canonical present | rel=canonical with a non-empty href | Search-engine adoption not judged |
| Matches final URL | Exact string equality | Trailing slash/parameter difference = no |
| og:url | meta property=og:url with content | Social rendering not judged |
| hreflang | Count of link hreflang | Target pages not validated |
Result: nine canonical tags, all self-referencing
| Page | canonical | Matches final | og:url | og = canonical | hreflang |
|---|---|---|---|---|---|
| E01 | no | — | no | — | 0 |
| E02 | no | — | no | — | 0 |
| E03 | yes | yes | yes | yes | 0 |
| E04 | no | — | no | — | 0 |
| E06 | yes | yes | yes | yes | 0 |
| E08 | no | — | no | — | 0 |
| E09 | yes | yes | yes | yes | 0 |
| E11 | no | — | no | — | 0 |
| E12 | yes | yes | yes | yes | 0 |
| E13 | no | — | no | — | 0 |
| E14 | yes | yes | yes | yes | 0 |
| E16 | no | — | no | — | 0 |
| E17 | yes | yes | no | — | 0 |
| E18 | no | — | no | — | 0 |
| E19 | no | — | no | — | 0 |
| E20 | no | — | no | — | 0 |
| E22 | yes | yes | yes | yes | 0 |
| E23 | yes | yes | yes | yes | 0 |
| E24 | yes | yes | yes | yes | 2 |
| E25 | no | — | no | — | 0 |
| E26 | no | — | no | — | 0 |
| E27 | no | — | no | — | 0 |
| E28 | no | — | no | — | 0 |
| E29 | no | — | no | — | 0 |
| E30 | no | — | no | — | 0 |
All nine canonical values are HTTPS, www-host, root-path and trailing-slash URLs matching the final URL; no cross-host, http-to-https or parameterised canonical appeared. The eight og:url values match canonical; E17 has canonical but no og:url.
Three ways to misread it
- No canonical does not mean no indexing. It only removes an explicit consolidation signal; search engines still use URLs, links and redirects. This study does not measure indexing.
- A matching canonical is not a clean bill of health. All 9 match, but the target was not checked for 200, noindex or sitemap agreement.
- og:url is not canonical. It is mainly social metadata; E17 has canonical without og:url, so canonical is not missing.
What this means for acceptance
- Write a homepage canonical. 16/25 have none; at least choose a primary URL for www/non-www and http/https variants.
- Compare exactly, not visually. Scheme, www, trailing slash, case and parameters must all match.
- Report canonical and og:url separately. The first is an indexing signal, the second social metadata; neither replaces the other.
- Check HTTP sites first. All four http homepages lack canonical; protocol and host duplication risk is higher.
- Require hreflang only when there are language variants. Only E24 has two entries; zero is not a defect for a monolingual site.
Limits
- 25-site pilot: the same homepage cache; not the industry and not all corporate sites.
- Cached HTML only: captured 2026-10-31; no JS execution and no runtime-inserted canonical.
- Exact strings: no normalisation of scheme/www/trailing slash/parameters; equivalent-but-different spellings count as no.
- No outcome test: indexing, ranking, duplicate consolidation and social cards are not measured.
- Homepages only: article, product and paginated pages are out of scope.
Data and reproduction
Anonymised data: equipment-homepage-canonical-og-2026.csv (25 rows × 6 columns: canonical present, canonical matches final, og:url present, og matches canonical, hreflang count). In-repo copy: docs/research/2026-11-homepage-canonical-og.csv.
Parser: Python regex over cached HTML for link rel=canonical and meta property=og:url, compared exactly with the captured final URL; zero network requests. Site names and full URLs are not published.
Related: the homepage resource inventory pilot, the homepage CSS/JS cache-header pilot, and the homepage image cache pilot.
Sources and definitions
- Sample: 25 homepage caches from study 4 (2026-10-31).
- canonical / og:url: static tags in cached HTML; final URL is the post-redirect URL at capture.
- Comparison: exact string; scheme, www, trailing slash, case and parameters are not normalised.
Frequently asked questions
Does no canonical mean no indexing?
No. It removes an explicit consolidation signal; search engines still use URLs, links and redirects. This study does not measure indexing.
Why are the four http homepages notable?
E04, E13, E16 and E27 were served over http:// and all lack canonical; explicit primary URLs matter more for protocol/host duplication.
Does a matching canonical mean the page is clean?
No. This study only compares strings; it does not check that the canonical target returns 200, is not noindex or agrees with the sitemap.
Can og:url replace canonical?
No. og:url is mainly social metadata; E17 has canonical without og:url, so canonical is not missing.
Is zero hreflang a problem?
Not for a monolingual site. Only E24 declares two entries; hreflang is needed when language or regional variants exist.
Can the data be reproduced?
Yes. The 25-row × 6-column anonymised CSV is in the repository and records canonical presence, final-URL match, og:url presence, og-canonical match and hreflang count; site names and full URLs are not published.