Nine of our pages are about a date on a calendar, and the first one we published is 353 days from being useful
We published 35 pages in one commit last night. Nine of them are about a specific date: Christmas, New Year, Burns Night, Valentine's, St Patrick's Day, Easter, Halloween, Bonfire Night, and a review of the year with the
We published 35 pages in one commit last night. Nine of them are about a specific date: Christmas, New Year, Burns Night, Valentine's, St Patrick's Day, Easter, Halloween, Bonfire Night, and a review of the year with the year in the URL.
Here is the distance from publication to the day each one is actually for:
| Page | The occasion | Days away |
|---|---|---|
| /quiz-questions/christmas | 25 December 2026 | 38 |
| /quiz-questions/new-year | 1 January 2027 | 45 |
| /quiz-questions/burns-night | 25 January 2027 | 69 |
| /quiz-questions/valentines | 14 February 2027 | 89 |
| /quiz-questions/st-patricks-day | 17 March 2027 | 120 |
| /quiz-questions/easter | 28 March 2027 | 131 |
| /quiz-questions/halloween | 31 October 2027 | 348 |
| /quiz-questions/bonfire-night | 5 November 2027 | 353 |
The ninth is a review of the year 2026, which has a window rather than a day: the fortnight either side of New Year, and then it is a page with last year's number in its URL.
Bonfire Night 2026 was eleven days before the page about it was written. So the page missed its occasion by a week and a half, and the next one it can serve is almost a year out.
That is not a scheduling failure. It is the only sensible move, and the reason is boring: a page that wants to rank for a seasonal query needs to have been crawled, linked and sitting still for months before the query volume arrives. Publishing a Bonfire Night page on 1 November is publishing it too late. Publishing it on 16 November, for next year, is early enough to be useless today and early enough to work.
The interesting part is what a site with no concept of a season does with pages like these.
Eight fields, none of them pointing at the future
Every public page is one node in one array, and this is every field those nodes use, counted across all 108 of them:
path, title, heading, shortLabel, description, cluster, updated, related
updated is the only date, and it means "the day the words last changed". It is a fact about the past. There is no season, no publishAt, no relevantFrom, no expiry. Nothing in the system knows that the Bonfire Night page is dormant for eleven and a half months of the year, or that one page's title contains a year that will stop being this year in six weeks.
That is a defensible place to start. 99 of the 108 pages are evergreen, and a field that nine pages use is a field eight people will forget to set. But the absence shows up in three places, all of which come out of that one updated string.
One field, three published claims
The sitemap. lastModified comes from the node, which was the fix for a sitemap that used to stamp every URL with the build date and tell a crawler the terms of service had changed this morning.
The visible line. 60 of the 108 pages pass an article flag to the page component, which renders a "Last reviewed" line under the heading:
Last reviewed 16 November 2026
The structured data. The same flag emits an Article, and both of its dates are the same field:
export function articleLd(node: ContentNode) {
const iso = `${node.updated}T00:00:00Z`
return {
'@type': 'Article',
// ...
datePublished: iso,
dateModified: iso,
}
}
Live, on the page that is 353 days early:
$ curl -s https://pub-trivia.app/quiz-questions/bonfire-night \
| grep -o 'date\(Published\|Modified\)":"[^"]*"'
datePublished":"2026-11-16T00:00:00Z"
dateModified":"2026-11-16T00:00:00Z"
The code has a comment explaining why those two are deliberately the same:
These are evergreen pages revised in place rather than dated posts, so the date
a guide was first written is not a fact about it that anyone needs; the date it
was last checked is.
I wrote that and I still agree with it for the guides. It is also the sentence that will become false, on a known date, for nine pages.
The lie that arrives next autumn
Suppose next October I do the right thing and re-read the Bonfire Night page before the season: check the facts, freshen a line, bump updated to 2027-10-20. Three claims move together:
- The sitemap says the page changed in October 2027. True.
- The visible line says last reviewed October 2027. True, and the whole reason the field exists.
- The
ArticlesaysdatePublished: 2027-10-20. False. The page was published in November 2026, which is the one date on it a reader or a crawler might genuinely want, because a seasonal page's value is partly that it has been around.
Today the two dates being equal is harmless, since every one of these pages has been published exactly once and never revised. The trap is that the first honest edit is what makes the markup wrong, so the code is correct until somebody does the right thing.
The clearest case against it is the page whose URL contains a year. A review of the year is not an evergreen page revised in place, it is a dated post, and a datePublished that moves every time it is touched describes something else entirely.
A collision that makes a wrong number look right
While counting, a nice coincidence. The sitemap has 111 URLs and seven distinct lastmod values:
$ curl -s https://pub-trivia.app/sitemap.xml | grep -o '<lastmod>[^<]*' | cut -c10-19 | sort | uniq -c
44 2026-09-14
1 2026-10-03
14 2026-10-15
1 2026-10-18
9 2026-11-04
4 2026-11-15
38 2026-11-16
38 URLs say 16 November, and only 35 pages changed on 16 November. The other three are the privacy, terms and cookie pages, which have no registry node, so they fall back to the build date. The build happened that day, so the fallback is indistinguishable from the real thing. Next week those three will say whatever day I last deployed, which is exactly the signal a crawler learns to ignore, and the only reason it looks fine right now is that the clock and the content agreed for one day.
What a season field would buy
The nine dated pages still link to each other, because related is hand written and a human wrote sensible lists: Valentine's points at Burns Night behind it and St Patrick's Day ahead of it. But Easter points at Bonfire Night, which is seven months behind it or five months ahead depending on which way you count, and nothing can tell the difference, because no code on the site knows when either of them is.
With a month or a date rule on the node, three things become possible that are not possible now. The questions hub could order itself by what is next rather than by the order the pages were written. A related block could carry one "coming up" slot that changes by itself. And a test could assert that no page whose path names a year is still linked once that year is over, instead of that being a thing I remember in January.
None of which is built. The point of writing it down is that the nine pages are live and the field is not, so the next person to add a seasonal page, which is me in about three weeks for Easter 2028, gets to decide whether nine is finally enough.
Have a look while they are all dormant
The questions hub lists all 35 category pages, and the dated ones are easy to spot because most of them are nowhere near their date. Open the Bonfire Night page and the "Last reviewed 16 November 2026" line is right there under the heading, saying the only true thing about a page's dates that the site currently knows.
Then curl the sitemap and count the lastmod values yourself. 44 of them still say 14 September, which is the correct answer for a page nobody has touched since, and the hardest signal in SEO to resist meddling with.
Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.