Pagination: the canonical mistake that hides pages
Paginated archives fail quietly. The pages all return 200, they all look right in a browser, and the products or articles on page two onwards stop appearing in search. The usual cause is a single line in the head of every page after the first.
The mistake
Pages 2, 3 and 4 declare page 1 as their canonical:
<!-- on /blog/page/2/ -->
<link rel="canonical" href="https://example.com/blog/" />That says: this page is a duplicate of page one, index that instead. It is a statement about the page, and it is false. Page two is not another copy of page one; it holds different items.
The cost is not the archive page itself, which nobody was searching for. It is everything only reachable from it. On a catalogue or a blog with a long back-file, the deeper pages are the only route to most of the site. Tell a crawler they are duplicates and it has little reason to keep fetching them, and the items behind them go with them.
Why sensible people do it
Because it looks like duplicate-content hygiene. The pages share a title, a heading and a layout, so collapsing them to one canonical feels tidy — and it is the shape of advice that circulated widely when pagination markup was a live topic. The instinct is right and the target is wrong: near-identical chrome around different items is not duplication.
What to do instead
Make every paginated page self-canonical. Page two's canonical is page two:
<!-- on /blog/page/2/ -->
<link rel="canonical" href="https://example.com/blog/page/2/" />Link the series properly in the HTML. Previous and next links, and numbered
links where the series is short enough, as real anchors a crawler can follow without running
JavaScript. This is what discovery now relies on: rel="prev" and
rel="next" are no longer used by Google for indexing, so the ordinary links are
doing the work.
Give each page something of its own where you reasonably can — a distinct title, or a heading that says which slice of the series it is. Not to rank the archive, but so nothing about the page invites the collapse you have just undone.
Checking it yourself, in one command
Fetch the second page of any series and read its canonical:
curl -s https://example.com/blog/page/2/ | grep -i 'rel="canonical"'If that prints page one, you have found it. Check a product category as well as the blog — these are usually set by a template, so the fault tends to be either absent or everywhere.
What this does not cover
It is about what the pages declare, not about whether pagination is the right design. An infinite scroll or a "load more" button that never changes the URL has the opposite problem — there are no deeper pages for a crawler to find at all, canonical or otherwise — and a view-all page is a different trade again. Those are decisions about the site. This is a line in the head that is saying something untrue about it.
Common questions
Should paginated pages canonicalise to page 1?
No. It tells search engines that pages 2 and beyond are duplicates of page 1, so they may stop being crawled — and everything only reachable from them goes with them. Each paginated page should be self-canonical.
Do I still need rel=prev and rel=next?
Google no longer uses them for indexing. Ordinary previous, next and numbered links in the HTML are what discovery relies on now, so make sure they are real anchors rather than JavaScript.
How do I check my pagination in one command?
Fetch the second page and grep its head for the canonical link: curl -s https://example.com/blog/page/2/ | grep -i canonical. If it prints page one, that is the fault. Check a product category as well as the blog, since templates make it all-or-nothing.
Is duplicate chrome on paginated pages a duplicate-content problem?
No. Sharing a title, heading and layout while listing different items is not duplication. Collapsing them with a canonical is the thing that causes harm.
What about infinite scroll or a load-more button?
That is the opposite problem: if the URL never changes there are no deeper pages for a crawler to find at all. It is a design decision rather than a declaration that is untrue, and needs a different fix.