Faceted Navigation and SEO: Controlling Crawl Without Losing Rankings
Faceted navigation lets users filter a catalogue by attributes like colour, size, brand, and price. It is great for shoppers and dangerous for crawling, because a handful of filters can combine into millions of URLs. The goal is to keep the filtered pages people actually search for and stop the rest from swallowing crawl.
Why facets are a crawl problem
Say you have five filter types with ten options each. The combinations run into the hundreds of thousands, and add sorting and pagination on top and it grows again. Most of those URLs are near-duplicates or thin, but they are crawlable and linked, so Googlebot finds them. On a large site this is the single biggest source of index bloat and wasted crawl.
Deciding what to index
The core decision is which filter combinations deserve to be indexable pages and which do not. Base it on search demand, not on what is technically possible.
| Filter combination | Index it? | Reason |
|---|---|---|
| Single high-demand attribute (waterproof jackets) | Yes | Real search demand, distinct intent |
| Two combined attributes with demand (women's waterproof jackets) | Often yes | Still a searched, distinct page |
| Sort and view variants (?sort=price) | No | Same content, no new intent |
| Deep multi-filter combinations | No | Thin, near-duplicate, no demand |
| Price-slider and range parameters | No | Effectively infinite, no demand |
Scroll table horizontally →
Do keyword research on the attributes. A filter that maps to real queries deserves a clean, indexable, self-canonical URL and internal links. Everything else should be controlled.
How to control crawl
There is no single switch. You combine a few mechanisms, each doing one job.
- Canonical tags point sort and parameter variants to the clean base URL, consolidating signals.
- Noindex keeps a filtered page usable for shoppers while removing it from the index.
- Robots.txt blocks genuine crawl traps that have no value and no rankings, such as tracking parameters.
- Crawlable versus non-crawlable links control which combinations Google even discovers, by rendering some filters in a way that does not create crawlable hrefs.
# robots.txt: block value-less parameters only
User-agent: *
Disallow: /*?sort=
Disallow: /*?view=
Disallow: /*sessionid=
# Keep demand-having facet folders crawlable
Allow: /jackets/waterproof/A key point people miss: robots.txt controls crawling, not indexing. If a blocked URL is linked, it can still be indexed without content. So pair crawl rules with canonical or noindex decisions rather than relying on robots.txt alone.
The mechanism you reach for should match the outcome you want. Use a self-referencing canonical when the filtered page should exist and rank in its own right. Use a canonical to the parent when the variant is essentially the same content in a different order. Use noindex when the page must stay usable for shoppers but has no place in search. And save robots.txt for the patterns that have no value to anyone, where you simply want to stop the crawler wasting requests. Mixing these up is how teams accidentally deindex pages they meant to keep, so decide the intent first and pick the tool second.
Keeping the pages that rank
The failure mode of a facet cleanup is de-indexing pages that quietly earned traffic. Before you apply any rule, pull the filtered URLs that currently get clicks and impressions from Search Console, and the ones with backlinks. Protect those. Make them self-canonical, keep them crawlable, and link to them from category pages so they read as intentional destinations rather than accidental facets.
Rolling it out
Do it in stages. Start with the clearest crawl traps, sort and tracking parameters, where the risk is near zero. Then handle the deeper filter combinations once you have confirmed the valuable ones are protected. Watch index coverage and log crawl distribution move over the following weeks. This is slow by nature, because Google has to recrawl and reprocess a large URL set.
How to tell it worked
Facet cleanup is slow and easy to second-guess, so decide up front what success looks like. The clearest signal is in the logs: the share of Googlebot requests going to sort, filter, and parameter URLs should fall, and the share going to real category and product pages should rise. That shift means crawl is being spent where it earns money, which was the whole point.
In Search Console, watch the indexed URL count move toward the set you actually want indexed, and keep an eye on the Indexed, not submitted in sitemap bucket shrinking. Most importantly, watch the traffic and rankings of the filtered pages you chose to protect. If those hold steady or grow while the total indexed count falls, you did it right. If a protected page slips, check that it stayed self-canonical and crawlable rather than getting caught by a broad rule.
Give it time. Google has to recrawl and reprocess a large URL set, so meaningful movement takes weeks, not days. Resist the urge to reverse course after a week of noisy data. Track the trend, not the daily wobble, and compare against the baseline you captured before the first change.
The takeaway: faceted navigation is not something to block wholesale or leave wide open. Decide what deserves indexing based on real demand, protect the filtered pages that rank, and control the rest with the right mix of canonical, noindex, and robots rules. On a large catalogue this is delicate work, and a crawl and indexation engagement exists precisely to get it right without losing rankings.