ECOMMERCE · SEARCH ARCHITECTURE
Keep the filters that help people shop. Give search engines only the collections you can justify and maintain as useful destinations—not every combination your catalogue can generate.
By TrendTransformers · 29 September 2026 · Based on current Google documentation. The office-chair example below is illustrative, not a client result.
A colour, size or price filter can be valuable after a visitor reaches your shop without deserving a separate place in search. The decision is not “filters on or off”. It is whether a particular selection answers a distinct buying need well enough to become a lasting landing page.
Promote one buying need, not a chain of filter clicks
Imagine an office-furniture shop with filters for seat width, back support, colour and price. A shopper chooses “compact”, “adjustable”, “grey” and a narrow price range. That selection is useful to that person. It does not, by itself, justify a new search page.
Now suppose enquiry records and relevant search queries repeatedly mention chairs for small home offices. A dedicated compact-office-chair collection may have a clearer job: help buyers compare footprint, usable seat dimensions and adjustment options. Before creating it, check whether an existing category already serves that same intent. Strengthen that page if it does.
The admission test is practical: can you name the buying problem, show a dependable selection that solves it, and explain the trade-offs better than the broad category? If the answer is only “we can generate this URL”, keep it as a filter. If demand is uncertain, treat one collection as a bounded experiment—not permission to create hundreds.
There is no universal minimum product count. Two genuinely different, reliably supplied options can be more useful than twenty near-identical listings. The question is whether a visitor arriving directly can make a sensible choice. Review stock continuity, product differences and the effort required to keep the page accurate.
This is an SEO website architecture decision before it is a plugin setting. Name the approved destinations first; implement their URL behaviour afterwards.
Write a small URL contract for the collection
For the fictional shop, the following is a proposed acceptance record—not a universal configuration to copy. The selected collection must remain recognisable when opened directly, without requiring the visitor to recreate the filters.
| URL or state | Intended job | Acceptance condition |
|---|---|---|
/office-chairs/ |
Broad browse category | Remains a useful entry point and links to the approved compact collection. |
/compact-office-chairs/ |
Selected search landing page | Shows the intended range, explains dimensions and trade-offs, and has a stable URL eligible for indexing. |
| Grey + narrow price band + newest-first | Temporary shopping state | Filters and sorting work for people; no separate search page is commissioned merely for this combination. |
| Impossible combination with no matching products | No valid result | Explains that no items match, offers a way to clear filters, and uses the correct HTTP response for that URL. |
For the approved page, align the address used in internal links, the sitemap and the canonical tag. Make it accessible to search engines, without a conflicting noindex instruction. Google’s ecommerce URL guidance recommends consistent URLs and ordinary linked navigation. Adding an eligible page does not guarantee that Google will index it.
Keep product-detail links separate from this collection decision. A link to an exact item must preserve the intended choice; the product-variant link test covers that different handoff.
Do not ask one setting to do three different jobs
Google’s faceted-navigation guidance warns that filter combinations can create excessive URL spaces and divert crawling from useful pages. The right response depends on the problem you have:
- Reduce unwanted crawling
- Control access to filter URL patterns that have no search role. Google’s guidance includes robots.txt rules for this purpose. Have the implementation checked against real URLs: an overbroad pattern can also block a collection you intended to expose.
- Keep a reachable page out of search
- Use an appropriate indexing instruction. Google must fetch the page to see
noindex; blocking it in robots.txt prevents that. Its noindex documentation explains why the two controls are not interchangeable. - Identify a preferred version
- A canonical tag signals the preferred URL among duplicate or closely similar versions. It is not an instant off-switch for crawling. Do not point an intentionally distinct compact-chair destination back to the broad category while also asking for the compact page to be indexed separately.
Do not paste a blanket block into a live shop to solve an indexing complaint. First determine which URLs already appear in search and which must remain eligible. Removing existing results and preventing future crawling require a deliberate sequence; they are not the same action.
For empty or nonsensical filter combinations, Google recommends an HTTP 404 response at the affected URL. That is not a reason to automatically delete a maintained category whenever stock temporarily runs out. Treat an invalid combination and a useful category with a replenishment plan as different cases.
Questions before changing product-filter indexing
Should every product filter have an indexable page?
No. Keep filters useful for shoppers, but select search landing pages only when they serve a distinct buying need and you can maintain a useful collection. A filter combination is not evidence of search demand.
Does blocking a filter URL remove it from Google?
Not necessarily. A robots.txt block controls crawling, not guaranteed removal from search. Google must be able to crawl a URL to see a noindex instruction.
Will more indexed filter pages improve AI-search visibility?
Not automatically. More URLs do not establish usefulness or guarantee citations. Prioritise clear, stable collections that answer a real buying need, then measure their contribution.
Commission a testable change, not “more SEO pages”
Give your developer one approved collection, the parent category, a normal filter state and an impossible combination. Ask for the actual URL, response status, index directive, canonical target and internal-link route for each. Then check the shopping experience on a phone: direct arrival, filter selection, clearing filters and opening a product must still work.
After release, compare the approved collection’s relevant search impressions, visits and useful shopping actions over a period that suits your traffic and seasonality. Also inspect whether unwanted filter URLs still dominate crawling. More indexed URLs alone is not success; neither is a cleaner crawl report if buyers can no longer find the right product.
If your existing setup already exposes a small, useful set of collections and keeps transient states under control, no mass rebuild is needed. If you cannot tell which rule applies to which URL, a technical SEO audit can establish that map before changes are commissioned.