Site taxonomy is the system of categories, tags and filters a site uses to group its pages, and on most platforms every term in it is also a page: a category archive, a tag archive, a filtered listing with a URL of its own. That makes site taxonomy a page-creation decision, and most of taxonomy SEO is deciding which of those pages should exist in the index at all.
The part of a site taxonomy people worry about most is the part Google cares about least. It doesn’t count the slashes in a URL. What matters more, by Google’s own account, is how many clicks a page sits from the homepage, and on most sites those are two different numbers.
Categories, tags and facets: three taxonomies, three kinds of pages
Most guides describe taxonomy types by shape: hierarchical, flat, faceted. The more useful description of a taxonomy is what it generates, because that is what search sees.
| Taxonomy | What it generates | Index it? | How it usually fails |
|---|---|---|---|
| Categories (hierarchical: parent, child) | One archive or listing page per category, often with pagination behind it | Yes, when the category matches a query people search | Named in the company’s vocabulary, so the page targets a query nobody types |
| Tags (flat: any number per item) | One archive page per tag, created the moment someone types a new tag | Rarely; only tags with demand of their own and enough items | A tag on one article, which makes a page that duplicates that article |
| Facets (filters: colour, size, brand, price) | A URL per filter combination, if the filters change the URL | Only the few combinations people search for by name | Thousands of combinations crawled, many of them empty |
Category names are where it starts. A taxonomy built for search uses the words searchers use: “compressor spare parts”, where the purchasing team would write “components for A-100 series equipment”. The second label can be correct and still produce a category page with no query behind it.
One distinction cuts across all three taxonomies. A filter that only rearranges what’s on the screen, with no URL change, creates no page and no SEO question. The moment a filter or a topic switch writes itself into the URL, as ?topic=seo or /insights/seo/, it becomes a page with everything that follows from that.
Every taxonomy term is a URL: which ones earn an indexed page
Jes Scholz put the core problem in one line in her guide on Search Engine Land: every new term auto-generates a new page, and that is where the SEO problems begin. An enterprise site may have 50 to 300 categories and subcategories and tens of thousands of pages generated by tags and facet combinations. Nobody decided to publish those pages. The CMS did, one taxonomy term at a time.
So the question for each term is the same one a pillar page asks of its subtopics, and it uses the same test. A term earns an indexed page when three things hold: people search for it as a phrase, the results for that phrase are listing pages or collections (which is what a term page is), and the term has enough items that its page says something no single item says. Searching the term and comparing its top ten with the top ten for its parent category is the check. If the results are the same pages, the term is a section of the parent, and the pillar pages guide walks through that test in detail.
Run across a whole niche, the same check can design a site taxonomy from scratch. Collect the keywords, group them by shared results, and the groups that stand apart are the categories; the ones that merge are one page. That is a sounder starting point than an org chart. It also settles the case for cross-category collections, the “coffee for the office” page that pulls grinders, machines and capsules from three categories. A collection like that is worth building when its phrase has its own results, and it is just another tag page when it doesn’t.
Terms that fail the check don’t have to disappear from the site. They can stay as navigation for people, without a page in the index. That is a decision about each term in the taxonomy, and one site-wide setting for all tags is a poor substitute for it.
Category pages: the landing pages for broad queries
In a hierarchical taxonomy the category page is usually the right URL for the broad query, and the most common category problem is a product or an article competing with it. Yoast’s example is swimwear: if every product page is optimised for “bathing suits”, the products compete with each other and with the category that should hold the term. The product pages take the specific queries, the category takes the broad one, and products link up to the category.
Whether a category page can hold the broad query depends on what ranks for it. Category pages win where the results are listings. Where the results are guides, the category page needs a written introduction that answers the query, or a guide that sits next to it and links in.
Platforms build their own taxonomy into the URL. WordPress puts categories under /category/ by default and tags under /tag/, and every post without a category lands in “Uncategorized”, which is a category page with no topic. Shopify lets a product be reached both through a collection, at /collections/{collection}/products/{product}, and at /products/{product}; the collection path canonicalises to the product path, so the index holds one URL, but internal links split between two addresses unless the theme links to the canonical one.
The same product or article often belongs in two branches of the taxonomy. Smart watches for running belong in sports and in fitness. The answer is to list the item in both categories and give it one URL, which most platforms do by default and some undo by putting the category in the item’s path.
Tag pages: when to index them, and why noindex isn’t a parking spot
John Mueller was asked whether tags help rankings and said: not necessarily. There is no magic around tags, in his words; they create more links and more pages within the site.
A tag page earns its place in the index on the same terms as any other taxonomy page, and most don’t: a tag on one article produces a page that is a list with one item, and a tag on three articles produces a thin page for a phrase nobody searches. Duplicates make it worse. Singular and plural versions of the same tag, or a tag with the same name as a category, give search two pages for one query. Scholz’s example is one TV show tagged six different ways on the same site.
The usual taxonomy SEO fix is to noindex every tag archive and keep them crawlable, so they keep passing links to the posts. That works for a while. Mueller said in December 2017 that a page on noindex for a long time is eventually treated as noindex, nofollow: Google stops following its links, and how long that takes depends on how often the page is crawled. A tag layer kept on noindex for years is a set of pages Google is gradually dropping, links included.
A page kept on noindex long term ends up treated like noindex, nofollow: Google eventually stops following its links (John Mueller, December 2017).
Tags that pass the check stay indexed and get an introduction. Tags that help readers navigate but fail the check go to noindex, knowing that noindex filters the index and, over time, stops passing links. Tags that do neither get merged into a better tag with a 301 or removed with a 410, which is the long-term fix Scholz recommends; her quick fix is an autocomplete in the editor, so writers pick existing tags instead of inventing new ones.
URL depth vs click depth: what Google counts
URL depth is the number of folders in an address. Click depth is the number of links a visitor follows from the homepage to reach the page. Both get filed under site taxonomy, they describe different things, and Google has said several times which one it reads.
| The claim | What Google has said | When |
|---|---|---|
| Fewer slashes make a page look more important | Mueller: “we don’t count slashes in the URLs”; what does matter “a little bit” is how easy the content is to find, meaning how many clicks from the homepage it takes | 2018, Webmaster Central hangout |
| A flat URL structure ranks better | Mueller: no need for an artificially flat structure; Google treats URLs as identifiers of content and doesn’t try to understand site structure from them | March 2021, Search Central hangout |
| Folder hierarchy is a ranking signal | Mueller: hierarchy in URL elements isn’t that important for Google search | February 2017, Twitter |
| Folders don’t matter at all | SEO Starter Guide: on a site with more than a few thousand URLs, grouping similar topics in directories can help Google learn how often the URLs in each directory change | Current documentation |
| Breadcrumbs show your hierarchy in the results | Since 23 January 2025 mobile results show only the domain; desktop results keep the breadcrumb | Search Central Blog |
The fourth row is the one people overcorrect against. Folders help crawling on large sites, because a directory that changes daily can be crawled more often than one that changes yearly. That is a crawling benefit on sites big enough to need it, and it has nothing to do with how a page ranks. Breadcrumbs still do their job on the page itself: each one is a link a level up, whatever the mobile results show.
Click depth is where a taxonomy does its work, and pagination is where it usually goes wrong. A listing that links only to the next page puts the hundredth post ten clicks deep at ten posts a page. Numbered pagination flattens that, until there are more pages than numbers shown. Categories, which are the taxonomy’s own answer, internal links from newer articles and links from service pages are what bring old material back up.
Two template settings make it worse. Some themes point the canonical of page two and beyond at page one, and Google’s pagination documentation says not to: each page in the sequence gets its own canonical URL. Others put noindex on every page after the first. Mueller has said that is acceptable. By his 2017 statement, though, the links on those pages eventually stop being followed, and for an old post the listing page is often the only internal link it has.
A “three-click rule” appears in several 2026 guides with Google Search Central named as the source: every important page within three clicks of the homepage. I couldn’t find that number in Google’s documentation. Mueller’s statements say closer is better and stop there.
What depth looks like on our own blog
Our insights section is small enough to show the whole taxonomy trade-off. At the time of writing it holds 26 articles, nine to a listing page, with numbered pagination and two rows of filters: topics (Design, Development, SEO, SEO Audit) and formats (Analysis, Guide, Note, Study). The taxonomy has no category archives. The newest article is two clicks from the homepage: the menu, then the listing. Four of the audit guides, among the earliest we published, sit on page three: menu, page three, article. Three clicks, because the pagination is numbered and every listing page is one click from the first. With next-only links it would be four, and it would grow with every nine articles published.
Three clicks is fine for 26 articles. A larger blog would ask whether to add a category page for “SEO audit”. This site already has a page with that name, the audit service page, and an article archive would compete with it for the same words. The check from earlier settles it. What comes back for “seo audit” is guides, checklists and tools, and an archive of a handful of articles is none of those. The better path to the audit guides is links from the service page and its sub-pages. Whether the topic filters are links to URLs or scripts that rearrange the listing decides whether they are already category pages; if they write to the URL, each of them is one, and each needs the same check.
Change the taxonomy, keep the URLs
Taxonomies change often. Categories go out of date, tags merge, a product line splits in two. URLs should change rarely, because every changed URL needs a redirect, and Yoast’s advice to existing sites is not to restructure URLs unless the benefit clearly outweighs that cost.
That argues for keeping the item’s URL free of the taxonomy. A post at /blog/post-name/ can move between categories without a redirect; a post at /category/subcategory/post-name/ can’t, and in WordPress, with the category in the permalink, it can change address when an editor changes the post’s primary category. The taxonomy pages themselves can carry the hierarchy in their own URLs. The items they list don’t have to.
Faceted navigation, briefly
Facets are the largest taxonomy problem on ecommerce sites and the topic of a document Google published in December 2024. Announcing it, Google said the most common source of the crawling complaints it gets is faceted navigation. The decisions are covered in our ecommerce SEO audit. Three of them run against common advice.
- To save crawling, block it. Google’s document puts robots.txt first for facet URLs that don’t need to be indexed. A noindex tag doesn’t save crawling, since Google has to fetch the page to read it.
- Return a 404 for empty combinations. A filter combination with no products should answer with a 404 status. A 200 page that says “no results” doesn’t do that, and Google’s document also advises against redirecting to a generic error page.
- Don’t pair noindex with a canonical to the parent. A popular rule sends deep filter combinations both signals at once. Mueller called noindex and rel=canonical contradictory pieces of information in 2018, and asked again in 2024 he said he would just pick one. In July 2026, about a different conflict, stock status that disagreed between a page and its product feed, he said there is no publicly defined order of precedence and that conflicting metadata should be fixed.
The indexable facets, the few combinations people search for by name, get a page treated like any other taxonomy page. Generating descriptions for thousands of them at once is the pattern Google’s May 2026 guide on generative AI features files under scaled content abuse when it’s done mainly to rank.
How to audit a site taxonomy
Start from the list of terms. Export every term in the taxonomy, categories and tags, with its item count, then pull each term URL’s clicks and impressions from Search Console with a page filter on /category/, /tag/ or the platform’s equivalent. A crawl adds the click depth of every item. The decisions come out of the three columns together.
| What you see | Where | Decision |
|---|---|---|
| A term with one to three items and no impressions | Term export, Search Console | Merge into a broader term with a 301, or remove with a 410 |
| Term pages reported as “Crawled – currently not indexed” in bulk | Search Console, Page indexing, filtered to the term folder | Google already declined them. Decide whether any deserve a rewrite; noindex or remove the rest |
| Two terms, or a term and a category, sharing impressions on one query | Performance, filter by the query, Pages tab | Keep the one with the stronger results and redirect the other |
| Page two and beyond of a listing on noindex, or with a canonical to page one | Crawler, indexability and canonical columns for /page/ URLs | Self-referencing canonical on every page of the sequence; drop the noindex unless the items have other internal links |
| Articles or products five or more clicks from the homepage | Crawler, click depth column | A category page, or links from pages that are already close to the homepage |
The rest is governance, and it’s where taxonomies decay. Someone owns the list of terms. New categories need a reason, ideally the check above. Neil Patel’s guide makes the point that the taxonomy should be written down: who owns updates, and when a term is added or retired.
Does site taxonomy improve rankings?
Most guides on taxonomy SEO say it does, and none of the ones in the current results cites evidence for it. What Google has said is narrower. Mueller has said that its quality algorithms look at the website overall, which is the reason thousands of thin term pages are worth cleaning up. Nobody outside Google has a threshold for how many such pages it takes before the site is affected, and I don’t have one either.
What a taxonomy does for certain is decide which pages exist and how many clicks separate them from the homepage. If a site’s taxonomy has drifted away from what people search for, our SEO audit starts there.