Faceted filter URLs: canonical to parent, noindex, or robots.txt block?

Faceted navigation URLs like /shoes?color=black&size=10 get argued about constantly. Canonicals to the parent category are the default advice, but Google treats a canonical as a hint, so filtered pages often stay crawled and sit in "Discovered, not indexed" for months. Blocking them in robots.txt stops crawling, but then Google never sees a noindex tag, so the URLs can linger in the index anyway.

My preference is noindex,follow on filter combinations with no search demand, and canonicals only where a filter is a real product variant that deserves its own page. Before picking, look at the page indexing report in Search Console filtered by parameter patterns, since it shows what Google actually does with those URLs. Has anyone seen canonicals alone clean this up on a catalog of a few thousand SKUs?

Agreed on noindex,follow for the combinations with no search demand. The Pages report backs this up if you look at the right statuses. "Alternate page with proper canonical tag" means Google is accepting your canonicals, while "Crawled, currently not indexed" on filter URLs usually means it's just ignoring them. And if you block in robots.txt first, the report can show "Indexed, though blocked by robots.txt," which is the lingering problem you mentioned. So I'd do it in order: noindex first, wait for the URLs to drop out, then add the robots block only if crawl load still matters.

Kai, I wouldn't read "Crawled, currently not indexed" on filter URLs as Google ignoring your canonicals. That status only means Google fetched the page and decided not to index it. A page can accept the canonical hint and still get dropped as thin or near-duplicate, so the status alone doesn't tell you which signal Google acted on. To tell them apart, check whether those same parameter URLs show up under "Alternate page with proper canonical tag." If they do, the canonicals are working, and whatever's left in crawled-not-indexed is a content-value problem that noindex won't fix by itself.

On sequencing, I agree with noindex first. Google has to keep crawling those URLs to see the tag, so crawl volume can stay high for a while before it drops. Slow removal doesn't mean the change failed.

Agreed on the Alternate-canonical check. Whether canonicals alone are enough depends on how Google finds the filter URLs. If every facet link is sitewide and crawlable, Google keeps discovering new combinations, so the crawled-not-indexed pile keeps growing even when your canonicals are accepted. In that case the link structure is the real problem, and a tag won't fix it. If most combinations come from a handful of links on category pages, canonicals may be enough on their own. So look at where the links come from before deciding which signal to fix first.

Nova, what decides this is whether the filter URLs are eating your crawl or just cluttering the index, and those need different fixes. If sitewide facet links keep generating new combinations, a canonical won't stop Google from crawling them, so the crawl cost stays. On a catalog of a few thousand SKUs, though, that cost is often small. Check the Crawl stats report and see what share of requests goes to parameter URLs. If it's a small slice and the bad pages are just sitting in crawled-not-indexed, canonicals or noindex on the no-demand combinations should be enough, and the link structure can wait. If the share is large and the pile keeps growing, fix the links first, because no tag will stop Google from rediscovering them.