Faceted Navigation: SEO for Online Store Filters

A “Laptops” category has ten brands, twenty colors, five memory options and several price ranges. For the shopper, filters make it possible to find the right model in a few clicks. For a search robot, those same filters can generate hundreds of thousands of addresses with almost identical content.

Faceted navigation is not a problem in itself — it is one of the most convenient things in e-commerce. The problem appears when a store does not distinguish which filter combinations are useful for search and which should exist only for the shopper’s convenience. Below is a step-by-step guide to putting this in order: from an inventory of parameters to rules for new filters. In 2024 Google updated its recommendations on faceted navigation, and most of the advice here is based on them.

How to tell that filters are already doing harm

The first symptoms are visible in Search Console. In the indexing report, the number of pages with the status “Discovered — currently not indexed” or “Duplicate without user-selected canonical” grows. In crawl stats, most robot requests go not to products but to addresses with parameters. New categories take weeks to get indexed, although it used to take days.

The second symptom is in search itself. For a brand query, Google shows not a clean category but a random filter combination with three products. The third is in server logs: every day the robot crawls thousands of sorting addresses that have never brought any traffic.

Step 1. Inventory of parameters

First, collect all address patterns the store can generate: filters by brand, by specifications, price, availability, as well as sorting, pagination, grid view switching and service tags from advertising.

Then calculate the combinatorics. A category with five filters of ten values each gives more than a hundred thousand theoretical addresses. It is usually this calculation that convinces the team that rules are needed. Mark which parameters change the set of products and which only change the order in which they are shown: sorting and grid view never need to be in the index.

Step 2. Split filters into three groups

In practice it is convenient to work not with each combination separately, but with groups. The first is landing filters: there is demand for them, the assortment is stable, and a separate page makes sense. The second is service filters: useful to the shopper but not to search. The third is junk combinations: three or more parameters, empty or nearly empty selections.

Check demand with data, not intuition — we covered how to do this in the article on keyword search volume. A benchmark for a landing page: at least several dozen products remain in the selection, and the query itself has stable demand throughout the year.

Step 3. Put the addresses in order

For landing pages, human-readable addresses such as /laptops/lenovo/ are more convenient. The rest of the combinations can stay on parameters, for example ?color=black. The main thing is that one selection has one address.

The most common technical mistake hides here: parameter order. If ?color=black&brand=lenovo and ?brand=lenovo&color=black open the same page, for the robot these are two different URLs, and it will crawl both. Parameters must be sorted by a fixed rule, empty ones removed, and unknown ones ignored rather than added to the address.

Step 4. Indexing signals: what to combine with what

For landing pages everything is simple: a self-referencing canonical, unique title and H1, a short description, links from the category and a line in the sitemap.

For service filters, the canonical points to the parent category and no regular HTML links lead to them. Junk combinations need noindex — and this is where the robots.txt trap lies. If you block an address in robots.txt, the robot will not open the page and will never see the noindex inside it, while already indexed addresses will stay in search without a description.

The correct order is this: first noindex and canonical, wait until the unnecessary pages leave the index, and only then, if the store is very large, limit crawling in robots.txt. Canonical does not control crawling — it is a hint about the preferred version, not a ban on visiting.

Step 5. Internal links and the sitemap

Google assesses the importance of pages through links. That is why landing filters should receive regular <a href> links from the category, the menu, “popular selections” blocks or blog articles. Service and junk combinations are better implemented with buttons or forms without separate addresses in the markup — that way they do not create new crawl paths.

The XML sitemap includes only canonical addresses the business wants to see in search. For large stores it is convenient to split it by type — categories, products, landing facets, articles. Then Search Console immediately shows which page type is indexed worse.

Empty pages: the most common cause of soft 404

The assortment changes, and a filter that showed twenty products yesterday may return none today. An indexed page saying “nothing found” damages the site’s quality assessment.

If the combination will no longer exist, returning a 404 is correct. If the page was deliberately merged with another, use a 301 to the relevant category. If the products will be back soon, a temporary noindex is enough. In any case, show the shopper alternatives without substituting the status code.

When filters run on JavaScript

Modern stores often update results without reloading the page. That is convenient but requires checking four things: the landing page opens via a direct link and shows the same set of products; the address changes in the browser history; the server returns the correct status code; the content is available after rendering.

The fastest way to check this is the URL Inspection tool in Search Console. If the store is moving to a new platform or a headless architecture, filter rules must be included in the migration map in advance — more on that in our plan for a redesign without losing SEO.

Five most common mistakes

The first is opening all combinations for indexing “to have more pages”. The second is blocking everything in robots.txt and expecting pages to disappear from the index. The third is pointing the canonical of one landing page to another landing page, so both lose a clear signal. The fourth is allowing indexing of sorting and pagination without need. The fifth is not writing the rules down, so every new attribute in the catalog spawns a new wave of duplicates.

The last mistake is the most expensive: without rules, the store returns to the original chaos within six months. That is why filters should be designed together with the product catalog structure — attributes determine which facets are possible at all.

What to monitor after the changes

For the first two weeks, watch the indexing report and crawl stats: the share of robot requests to parameter addresses should fall, and requests to products and categories should grow. Then look at 404s so you do not remove too much, and at the positions of landing facets. After two or three months, assess the main thing: impressions, clicks and sales from filter pages.

Frequently asked questions

How many filters can be opened for indexing?

There is no universal number. Usually these are brands, one or two key specifications and popular combinations of two parameters. Combinations of three or more filters are indexed rarely and only with confirmed demand.

Which is better — parameters or clean URLs?

For pages you promote, human-readable addresses are more convenient. For the rest, parameters are perfectly fine if you manage them with consistent rules.

Do filters harm SEO?

Not by themselves. Harm comes from uncontrolled combinations, duplicates and empty pages. With the right rules, facets give a store dozens of additional landing pages for specific queries.

How long does it take to put this in order?

The audit and the rules usually take one to two weeks. Technical implementation depends on the platform. Results in the index are visible within one to two months after the changes.

Faceted navigation should help the shopper and at the same time avoid creating chaos for the search engine. The GL.ua team sets filter rules at the stage of online store development, so that SEO does not turn into endless address fixing after launch.