Direct answer: Put three “freshness signals” side by side and they often disagree. 15 of 25 homepages expose a copyright year, but only 3 say 2026; 11 of the 13 usable sitemaps carry a 2026 lastmod; 13 homepages contain a 2026 date string. On 3 sites the sitemap is already 2026 while the footer still says 2017 or 2023. A copyright year is not evidence that content was updated.
This study fetches nothing new. It parses existing caches only: the 25 homepage HTML files from study 4 and the raw sitemap files saved in study 10. The question is whether a common inference — “the footer says this year, so the site is maintained” — survives contact with the data.
Method: zero requests, existing caches only
Homepage HTML comes from the study-4 cache (crawled 2026-10-31 to 11-01) and sitemaps from the raw files saved in study 10 (2026-11-01 20:41–20:46). This study makes no network request; it parses static HTML and XML only and runs no JavaScript.
| Field | Definition | Boundary |
|---|---|---|
| Copyright year | A 20xx within 60 characters after © / © / Copyright | First match only; 0 when absent; avoids model and order numbers |
| lastmod | Lexicographic maximum of all lastmod values in the site's sitemap | ISO dates compare directly; sites without a sitemap are excluded |
| Homepage date | Whether the HTML contains 2026-, 2026/ or 2026年 | Proves a string exists, not that content changed in 2026 |
| Signal conflict | lastmod in 2026 while the copyright year is earlier | Describes the difference; does not explain it |
Result: the three signals disagree
| Signal | Hits | Denominator | Distribution |
|---|---|---|---|
| Copyright year | 15 | 25 homepages | 2026: 3; 2023: 4; 2025: 2; 2017: 2; 2024, 2022, 2011, 2002: 1 each; 10 homepages had no match |
| Sitemap lastmod | 13 | 24 sitemaps (13 usable) | 2026: 11; 2025: 1; 2024: 1 |
| Homepage 2026 date | 13 | 25 homepages | Any 2026 string, including news lists, events and specifications |
The copyright denominator is 25 homepages; the lastmod denominator can only be the 13 sites with a usable sitemap. The two counts should not be subtracted from each other.
Three sites: sitemap says 2026, footer does not
| sample_id | Copyright year | Latest lastmod | 2026 date on homepage |
|---|---|---|---|
| E01 | 2017 | 2026-09-04 | yes |
| E02 | 2023 | 2026-10-28 | yes |
| E14 | 2023 | 2026-10-28 | yes |
These rows show a disagreement between two fields, not that the site is stale or the data is fake. At least three ordinary explanations exist: the sitemap is generated by a CMS and updates with content while the footer is hard-coded; the copyright marks first publication; or the year is maintained by hand and simply not touched. The reverse also holds — a 2026 copyright can be a template that increments itself every year.
Why a copyright year is not update evidence
- Hard-coded template: a footer is usually a global component, changed once and kept for years.
- Automatic increment: some CMSs render the current year regardless of content.
- A different legal meaning: copyright marks a claim or first publication, not the last edit.
- Often missing: 10 of 25 homepages had no copyright match at all.
What this means for acceptance
- “Recently updated” needs a verifiable record. Footer years and homepage dates are weak signals and cannot stand alone as delivery evidence.
- lastmod is a file field. It proves what the sitemap says, not that the page changed that day; it only helps when aligned with real changes.
- Read the three signals together. Agreement raises confidence; disagreement should defer to evidence that can be re-checked.
- Agree the parsing rules up front. Where the copyright is read, whether lastmod is max or per URL, which date formats count.
- Separate “has a date” from “has an update”. The first is a string match; the second needs version records or a content diff.
Limits
- Cached, zero-request: homepage and sitemap captures happened at different times and do not describe one instant.
- Incomplete coverage: 10 homepages had no copyright match; 13 sites had no usable sitemap.
- Regex boundary: the copyright window is 60 characters, so separated wording can be missed.
- Weak signal: “2026 date” includes any 2026 string and cannot separate events, news and body copy.
- Non-random sample: 25 companies, a pilot; no claim about the industry.
Data and reproduction
Anonymised data: equipment-freshness-signals-2026.csv (25 rows: sample_id, copyright_year, lastmod_max, lastmod_year, home_2026_date, stale_footer). The repository keeps docs/research/2026-11-freshness-signals.csv.
Parser: scripts/freshness_probe.py; unit tests: scripts/test_freshness_probe.py (14 tests, no network). Run:
python -m unittest scripts.test_freshness_probe
Related: robots scan of 30 sites, sitemap scan of 24 sites (the lastmod source), the 12/24-hour recheck (change counts for the same URLs), and the freshness checker (these rules as a local browser tool).
Sources and definitions
- Homepage HTML: the 25 cached homepages from study 4 (crawled 2026-10-31 to 11-01, UA
LingyiResearchBot/1.0). - Sitemaps: raw files saved in study 10 (2026-11-01 20:41–20:46); 13 were parseable.
- Parsing: 2026-11-02, zero network requests; static HTML/XML only, no JavaScript.