Skip to content
search-engine-optimizedby Netsolution

Pagination and faceted navigation

A pagination is walked through real links, and each numbered page carries its own address and its own canonical. Faceted navigation, by contrast, manufactures an immense number of worthless addresses that crawlers visit before they can judge how useless they are.

Two problems people conflate, with different degrees of severity

Pagination cuts a list into numbered pages: the number of addresses produced is known and proportional to the content. Faceted navigation crosses filters — size, colour, price, brand, sort order — and the number of possible addresses is the product of the options. It is astronomical by construction, and it says nothing about the real quantity of content.

Hence the order of this guide: pagination is a matter of tidiness, faceted navigation a matter of volume. Treating them with the same gestures is the first mistake.

The documentation asks that each page include a link written with a link tag to the next one: this helps the crawler find the following pages.

officialSource : Google Search Central, “Pagination, incremental page loading, and their impact on Google Search”

A real link, that is, a link tag carrying an address, and not a button that fires a script. The distinction is not aesthetic: what is not a link is not a path, and a page reached by no path does not exist to a crawler.

Every page in a sequence must carry a unique address, for instance by means of a page number parameter: the addresses of a sequence are treated as distinct pages.

officialSource : Google Search Central, “Pagination, incremental page loading, and their impact on Google Search”

The second half of that sentence is the more useful one: these addresses are treated as distinct pages. They are therefore not views of a single page, and anything done to group them works against access to the content they carry.

The first page of a sequence must not serve as the canonical for the others: each page gets its own canonical address.

officialSource : Google Search Central, “Pagination, incremental page loading, and their impact on Google Search”

This is the most useful prohibition in the document, because it targets a practice still widely advised elsewhere. Declaring the first page as the canonical for the whole sequence amounts to asking that only the first screen of a list be indexed. On a catalogue, that means hiding the bulk of your own products.

Fragment identifiers, the part of an address after the hash, must not be used as a page number: the engine ignores them.

officialSource : Google Search Central, “Pagination, incremental page loading, and their impact on Google Search”

That sentence also settles, in passing, the fate of progressive loading arrangements: what exists only after the hash does not exist. Infinite scrolling is not forbidden, but it has to rest on genuine numbered addresses, without which the content past the first screen has no path.

Giving every numbered page its own address and its own canonical

hypothesis

Effort
moderate
Time to effect
weeks
Risk
none
Durability
structural
A nineteenth-century patent plate showing columnar registers: the same deeds are filed once by claimant then by defendant, once by grantor then by grantee.
The same records, reordered for every way of searching them. Convenient for whoever consults them, and the price is visible on the plate: each new order copies the whole register again.George C. Ward — Domaine public · Wikimedia Commons, relevé le

Faceted navigation: a volume problem, described by the engine

Addresses created by faceted navigation look new and crawlers cannot know whether they are useful without having crawled them: they therefore visit a very large number before concluding they serve no purpose.

officialSource : Google Search Central, “Managing crawling of faceted navigation URLs”

That description is worth reading slowly, because it contains the whole mechanism. Crawlers cannot know in advance that a filter address is of no interest. They learn it by visiting. The cost is therefore paid before the judgement, and it is paid on every combination. No quality of content compensates for that, since the problem precedes any reading of the content.

Crawling spent on useless addresses is time crawlers do not spend on new and useful ones.

officialSource : Google Search Central, “Managing crawling of faceted navigation URLs”

There is the consequence you see in the real life of a site: new pages are slow to be discovered, and nobody connects that delay to the filters. It is the same reasoning as crawl budget, applied to its most frequent cause.

The two documented steps on facets

The documentation asks that the industry-standard parameter separator be used: commas, semicolons and brackets are hard for crawlers to recognise as separators.

officialSource : Google Search Central, “Managing crawling of faceted navigation URLs”

A detail of form, a real effect: a crawler that does not recognise the boundary between two parameters cannot recognise two addresses as two variants of the same list. “Creative” separators — commas, semicolons, brackets — are therefore paid for in needlessly unique addresses.

A filter combination that returns no result must answer with an error code stating that the resource does not exist.

officialSource : Google Search Central, “Managing crawling of faceted navigation URLs”

In other words: a filter combination with no results must not answer with a page that exists and is empty. A catalogue that politely answers “no product matches” to any combination manufactures an unlimited number of perfectly valid empty pages — and every one of them will be crawled.

As for the choice of blocking tool, the documentation ranks them without ambiguity: disallow crawling of those addresses when you do not need them indexed, and treat the canonical as a slower means, which may reduce the crawl volume of variants over time. The first acts on the cause, the second on its consequences.

What does not work

Point a canonical from all your filtered and paginated listing pages to the first page of the category: that way you concentrate the power on a single address.

The documentation explicitly refuses using the first page of a sequence as the canonical for the following ones, and asks for a canonical per page. The advice quoted mixes two distinct cases — pagination and filters — and applies to the first a remedy that, for it, closes access to the content.

One last point, often handled backwards: the solution is almost never technical first. The question to ask before any setting is which filtered lists deserve to exist as pages, that is, which of them answer a demand real readers have formulated. The answer is usually a small number, to be treated as pages in their own right; all the rest is a display state, and a display state does not need a crawlable address.

Sources

  1. Pagination, incremental page loading, and their impact on Google Search — Google Search Centralaccessed on
  2. Managing crawling of faceted navigation URLs — Google Search Centralaccessed on
  3. How to specify a canonical URL with rel="canonical" and other methods — Google Search Centralaccessed on

Block revalidated on

Frequently asked questions

Should the first page be declared canonical for the rest?
No, the documentation explicitly refuses this and asks that each page get its own canonical. The reason is guessable: the pages of a sequence do not carry the same content, so grouping them amounts to asking the engine to forget everything not on the first.
Are the next-page and previous-page tags still useful?
The documentation states that this engine no longer uses them, while noting that other engines may still do so. What counts now sits elsewhere: real links, written in HTML, from one page to the next, which the crawler can follow.
Is infinite scrolling a problem?
It rests on script, and therefore on an extra engine-side stage, and it often exposes no proper address per page. The documented remedy is not to remove it but to back it with genuine numbered addresses, reachable through links.
Should filters be blocked in the robots.txt file?
That is the option the documentation recommends when you do not need those addresses indexed. The canonical stays possible but acts slowly, reducing crawl volume over time. The two therefore do not address the same urgency.