Paginated series, and the depth they quietly create

With only a next control, the shortest route to the last page of a series runs through every page before it. That is the entire problem, and it is arithmetic rather than a judgement.

Numbered links change the arithmetic immediately. A control showing first, previous, a window of nearby numbers, next and last puts every page in a forty-page series within two or three hops: straight to page 40 from the last link, or to page 20 and onward. The same content, the same URLs, and the depth problem is gone.

<!-- one link forward. page 40 is 39 hops from here -->
<a href="/blog?page=2">Next</a>

<!-- a window plus the ends. every page is within a few hops -->
<nav aria-label="Pagination">
  <a href="/blog?page=1" aria-label="First page">First</a>
  <a href="/blog?page=17" rel="prev">Previous</a>
  <a href="/blog?page=16">16</a>
  <a href="/blog?page=17">17</a>
  <span aria-current="page">18</span>
  <a href="/blog?page=19">19</a>
  <a href="/blog?page=20">20</a>
  <a href="/blog?page=19" rel="next">Next</a>
  <a href="/blog?page=40" aria-label="Last page">Last</a>
</nav>

Two accessibility details in that markup are worth keeping. The container is a nav with a label, so the control is identifiable and skippable, and the current page is marked with aria-current rather than being styled and left ambiguous. Both are free, and the first is part of the same argument about marking regions for what they are.

The failure that beats all of these is a control that does not exist in the served HTML. A paginator rendered in script has no links to follow, so the series has exactly one crawlable page, which makes this a dead-end route rather than a depth problem.

The first: canonicalising every page to page one. This is the most damaging thing you can do to a series and it is done constantly, because it looks like duplicate-content hygiene. Page 2 is not a variant of page 1; it contains different items. A canonical pointing at page 1 asks for page 2 to be dropped, and with it every item that only appears there. Each page in a series should be canonical to itself.

The second: one title for the whole series. Forty pages all called Blog | Fielding Cutlery is the largest single source of duplicate titles on most sites. Adding the page number is a one-line template change and it fixes a finding in a different pillar as a side effect. The same applies to the meta description, where the honest answer is often to omit it on pages 2 and beyond rather than to repeat one.

Two smaller decisions worth making deliberately. Whether page one has a parameter at all: /blog and /blog?page=1 serving the same list is a duplicate pair, so pick one form and redirect the other. And whether paginated pages should be indexable: keeping them indexable is the default and the right choice when they are the route to your content, since a noindex on them eventually stops the links being followed.

Yes, and mostly as a consequence rather than as a subject. Nothing about pagination is machine-readable in a way that helps retrieval. What matters is that the items in your series are reachable, and a next-only or script-rendered paginator means most of them are not.

The specific loss is worth naming: on a content site, the deep pages of an archive hold the older material, which is frequently the most thorough and the most specific. An agent that follows two links from a search result reaches page one and nothing else, so the best answer you have written may be structurally unavailable while a competitor's shallower page is not.

There is a secondary effect from the title problem. Forty pages that describe themselves identically are forty documents nothing can tell apart, and a system deduplicating a corpus has to choose. That is the same mechanism clustering describes, arrived at through a paginator rather than through a print view.

Two questions: are the page links in the HTML, and does each page declare itself canonical.

LIST=https://example.com/blog

# are numbered links in the served markup at all?
curl -s "$LIST" \
  | grep -oE 'href="[^"]*(page=|/page/)[0-9]+[^"]*"' \
  | sort -u

# walk a few pages and check the title and canonical on each
for n in 1 2 3 40; do
  url="$LIST?page=$n"
  body=$(curl -s "$url" | tr '\n' ' ')
  title=$(printf '%s' "$body" | grep -o '<title>[^<]*' | sed 's/<title>//')
  canon=$(printf '%s' "$body" \
    | grep -o 'rel="canonical" href="[^"]*"' | sed 's/.*href="//;s/"$//')
  printf 'page %-3s title: %-40s canonical: %s\n' "$n" "$title" "$canon"
done

Three failures show up immediately in that output. Identical titles across the four rows is the duplicate-title problem. Every canonical pointing at page one is the mistake that removes your series from the index. And an empty result from the first command means the paginator is not in the HTML, which is the serious version.

Then check page 40 exists at all. Series that were generated from a count sometimes render an empty page past the end rather than a 404, which quietly adds an unlimited number of empty pages to your site.