Structure URLs and internal links

Which Faceted URLs Should Google Index—and How Do You Limit the Rest?

Decide which faceted category URLs deserve landing pages, then test links, canonicals, noindex signals, and 404 responses by URL pattern.

Editorial illustration for Which Faceted URLs Should Google Index—and How Do You Limit the Rest?

A filter is useful to a shopper without necessarily being useful as a search landing page. The practical decision is to select the filtered category pages you want Google to find, then define what happens to every other pattern of filter URL. You do not need to review every possible combination individually.

Map the patterns before changing controls

Consider a hypothetical shoe category. These URLs illustrate four different jobs:

URL Intended role
https://www.example.com/shoes/ Parent category
https://www.example.com/shoes/?brand=acme Possible brand landing page
https://www.example.com/shoes/?brand=acme&color=red&size=9 Narrow, nonempty shopping state
https://www.example.com/shoes/?sort=price_asc Sort order, not a separate landing page

This is an example policy, not a rule that every brand filter deserves indexing. Inventory the filters, their options, and the URLs the interface actually generates. Include repeated selections, empty results, and pagination beyond the last page. A small category can still produce a large URL space when filters combine.

That matters for discovery, not just tidy reporting. Google explains that crawlers may visit many apparently new faceted URLs before determining they are unhelpful; time spent there can slow discovery of useful URLs (Google’s faceted-navigation guidance).

Choose landing pages on their own merits

For each candidate, ask whether people seek that category-and-filter combination, whether it has enough results to serve them, and whether the page differs meaningfully from its parent and neighboring facets. Search Console impressions and clicks can inform the demand question, but an absence of impressions is not, by itself, a verdict on a page Google has not found. The faceted-navigation analysis from cognitiveSEO presents demand, supply, and distinctiveness as selection criteria—not indexing thresholds.

In the example, a well-stocked Acme shoes page could be worth maintaining as a landing page. An Acme, red, size 9 selection may still be valuable while shopping, but if it largely repeats other pages or has few results, it need not become another search landing page. Revisit that choice when the catalogue changes. Do not turn an empty selection into a thin landing page: Google recommends a 404 when the combination has no results.

Give each chosen page one URL and a link

Pick a parameter order and generate it regardless of the order in which someone clicks filters. In this example, the order is brand, color, then size, joined with &. Do not also emit ?size=9&brand=acme&color=red as an alternative route to the same state, and reject duplicate filters. Google specifically recommends & for parameters and, when filters are encoded in paths, a consistent logical order with no duplicate filters.

For the chosen brand page, publish a normal HTML link from a relevant category page. Its response should be HTTP 200, and its canonical should point to that same URL:

<!-- On the shoes category page -->
<a href="https://www.example.com/shoes/?brand=acme">Acme shoes</a>

<!-- In the head of the Acme shoes page -->
<link rel="canonical" href="https://www.example.com/shoes/?brand=acme">

A JavaScript-only control should not be the sole route to a page you intend crawlers to discover: important facet links need crawlable anchors. Nor should a fragment such as #brand=acme be its only distinct address. Google says fragment-based filters have no effect on crawling, positive or negative; that is different from providing a linked, distinct landing-page URL.

Apply different policies to valid states and invalid URLs

Keep the controls specific to the pattern. If the sort URL displays the category’s products in a different order but is not a chosen landing page, this illustrative canonical points it to the parent. Place it in the head of the sort page, not the parent:

<link rel="canonical" href="https://www.example.com/shoes/">

For a nonempty narrow combination that should remain available to shoppers but is not a landing page, a separate policy could publish this in that combination page’s head:

<meta name="robots" content="noindex">

Do not apply both examples to the same pattern. A canonical expresses a preferred version and may reduce crawling of noncanonical versions over time; it is not a guaranteed crawl stop, and Google may choose differently. The facet-control discussion from Product Lasso likewise cautions against combining canonical and noindex on one URL.

An invalid URL needs a response, not an indexation hint on an otherwise successful page. If ?brand=acme&color=red&size=9 has no products, return HTTP 404 at that requested URL. Do the same for duplicate or nonsensical filters and nonexistent pagination URLs. Do not redirect every empty selection to a shared error URL; Google’s guidance explicitly calls for the 404 at the URL requested.

Test the policy by pattern

Take several URLs from each group: chosen landing pages, sorts, nonempty narrow combinations, empty combinations, duplicate filters, and nonexistent pagination. Follow the links your category page publishes and inspect the generated URLs. Then check each response status, whether robots rules permit access where you expect Google to read a page signal, and the canonical or noindex actually published in its HTML. A chosen page should resolve at its one consistent URL with HTTP 200, a crawlable incoming link, and a self-referencing canonical. Invalid samples should return 404 at their own URLs.

A passing implementation check establishes what the site serves, not what Google has indexed. Compare later crawls and Google’s indexing observations by URL pattern: are unwanted variants still being linked or found, and which URL is Google treating as indexed? Crawl-volume changes and canonical selection need observation over time. If the canonical itself is wrong, check the published URL before interpreting an indexing report.

Find a note

Search by topic, title, or keyword.