Website architecture is the way pages are grouped and linked to each other: which pages exist, which sections they sit in, and how many links it takes to get from one to another.
The part that matters to Google is the second. Google Search Central's guide for e-commerce sites says that "Google generally doesn't look at the structure of URLs to work out the structure of a site", and that it looks at the links between pages instead. Anyone redesigning the site tree starting from the URL folders is working on the wrong side of the problem.
Site architecture, site structure, site hierarchy: three names for the same thing
They are used interchangeably, and that is fine. "Hierarchy" stresses the tree diagram, with the home page at the top and then categories, subcategories and end pages; "architecture" also covers the decisions you can't see in the diagram. They refer to the same object.
One distinction that does matter: the navigation menu is not the site structure. The menu shows the items you have chosen to highlight, the structure includes every page on the site, including those you can't reach from the menu. On an e-commerce site with four thousand product pages the two don't even look alike.
What Google's documentation says, and from what threshold
Google infers the relative importance of a page from two signals: how many links it has to follow to reach it and how many links point to it. It is in the same e-commerce guide, and it is why depth really matters. Not as a threshold to respect, but as a cost of discovery.
The SEO Starter Guide, updated in December 2025, sets an explicit threshold: how you organise your content may affect crawling and indexing "if you have more than a few thousand URLs on your site". Below that number, the same page adds, search engines will likely understand your pages as they are, and there is no need to "drop everything and start reorganizing your site". It is a sentence no agency has an interest in quoting, and it sits on the most-read page of the documentation.
The same goes for crawl budget, which turns up in almost every article on site structure. Google's guide to crawl budget, updated on 22 July 2026, states at the top who it is for: sites with more than a million pages whose content changes about once a week, or sites with more than ten thousand pages whose content changes daily. A two-hundred-page company website doesn't qualify, and its architecture should be decided for other reasons.
On Google UK the query "website architecture" (390 searches a month) opens with an AI Overview that cites isev.co.uk, Semrush, Terakeet, NordVPN, Figma and a few others, and lists URL structure among the key components because it "clearly shows the path and relationship of a page within the site". None of the sources it cites is developers.google.com, and the organic results below it mix luxury web design opinion pieces with architecture firms' websites (DataForSEO, Google UK, September 2026). The rules that get repeated come from somewhere else.
The three-click rule never had a source
The rule first appears in print in Jeffrey Zeldman's Taking Your Talent to the Web, 2001, where it is taken for granted with no study behind it: if users can't find what they want within three clicks, they "might" go elsewhere.
In 2003 Joshua Porter of User Interface Engineering went and checked. He took the data from a study of 44 users and 620 tasks, more than eight thousand clicks counted one by one, and found no correlation between the number of clicks and task success. People who gave up after three clicks were no more numerous than those who gave up after twelve, and 80% of tasks were only completed by the fifteenth click. Stated satisfaction didn't move either: the share of dissatisfied users ranged between 46% and 61% whatever the length of the path. It is a US sample from 2003, and no published replication exists.
The Nielsen Norman Group came back to it in 2019 with an observation worth more than the rebuttal: the three-click rule and the rule of seven top-level menu categories both lack data, and they contradict each other. To stay within three clicks you have to widen the first level; to stay within seven items you have to go deeper. Follow both and you end up choosing between two mediocre experiences.
| Rule in circulation | Where it comes from | What the documentation and data say |
|---|---|---|
| Every page no more than three clicks from the home page | A 2001 web design book, with no data | Across 620 measured tasks there is no correlation between clicks and success (UIE, 2003, US sample) |
| Seven or eight main categories at most | A usability rule of uncertain origin | The Nielsen Norman Group classes it as a myth, and in conflict with the three-click rule |
| URL folders define the structure | Common agency practice | "Google generally doesn't look at the structure of URLs to work out the structure of a site" |
| Good structure saves crawl budget | General-audience articles | The crawl budget guide is aimed at sites with over 1,000,000 pages, or over 10,000 with daily updates |
| Architecture is a ranking factor | Simplification | Google talks about relative importance inferred from inbound links and distance from the home page, not a score |
Where architectures really break: filters
On an e-commerce site the damage is never done by the fourth menu level. It is done by filters. Google's documentation on faceted navigation, updated on 6 March 2026, is blunt: the most common implementation, based on URL parameters, can generate infinite URLs. Three filters with ten values each are enough to turn a category of a hundred products into thousands of different addresses.
The two harms Google lists follow from that: overcrawling of URLs that will turn out to be useless, and slower discovery of new pages, because the crawler has spent its time elsewhere. It is the crawl budget problem coming back in through the window, but this time because of a design mistake rather than the size of the site.
The countermeasures, in the order of effectiveness the same page gives:
- prevent crawling of filtered URLs if you don't need them indexed, leaving only the unfiltered category page and the individual product pages crawlable. It is almost always the right choice;
- move the filter mechanism to URL fragments, the part after the hash, which has no effect on crawling, positive or negative;
- if some combinations need to stay indexable, use the ampersand as the parameter separator (crawlers struggle to recognise commas, semicolons and square brackets) and always keep the filters in the same order in the path;
- rel="canonical" and rel="nofollow" are still possible, but Google describes them as less effective in the long run, and nofollow only works if it is on every single link pointing to that URL.
How to design an architecture that holds up
Categories come from queries, not from the organisation chart. Two products the sales team considers the same family belong on two different pages if they answer two different search intents, and two separate departments can easily end up on the same page.
Then there are three checks worth more than any well-drawn diagram:
- every page you want to rank must receive at least one internal link from a page Googlebot already visits. A page that exists only in the sitemap is discovered late;
- links must be a href tags. Google explicitly asks you not to rely on JavaScript events on other DOM elements for navigation, and answer engine crawlers are even less forgiving than Googlebot;
- if a product can only be reached from the internal search bar, it doesn't exist for Google: Googlebot generally doesn't submit searches into search boxes while crawling. Category pages must link to every product page, otherwise sitemaps or feeds have to cover the gap.
Breadcrumbs are the cheapest way to shorten the distance on every deep page, and Google also uses them to build the path shown in search results. URL structure, on the other hand, remains a matter of readability for people: useful, but it is not where the search engine reads the hierarchy.
Changing the architecture without losing what works
A reorganisation is almost always a URL operation, and there are three rules. A one-to-one 301 redirect from the old page to the new one, never a mass redirect to the home page. No redirect chains, which the crawl budget guide describes as harmful to crawling. A 404 or 410 for what really disappears, because it is a strong signal not to come back to that address, whereas a blocked URL stays in the crawl queue much longer.
Before touching anything, export from Search Console the pages that received clicks in the last sixteen months, and keep them aside. After the migration you will compare the same pages. It is the only way to know whether the work succeeded, and it has to be prepared beforehand, because afterwards the data has already changed. On a site with established traffic it is the kind of job we handle within SEO consulting, precisely because mistakes cost months.
Something rarely said in the industry: the three-click rule has survived twenty years of rebuttals because it is the only one you can check in a meeting. Anyone can count clicks, even someone who has never opened Search Console. Judging whether the menu item "Solutions" tells visitors what they will find behind it is something nobody can do at a glance, and so it doesn't get discussed. If you have half a day to spend on architecture, rewrite the menu labels before moving a level: an opaque label loses more visits than an extra click.
Frequently asked questions about website architecture
There is no right number: Google's documentation contains no click or level threshold. It talks about relative importance inferred from inbound links and the number of links to follow to reach a page. In practice, keep the pages that bring in revenue close to the home page, leave support pages deeper, and check that no useful page is left without inbound links.
The menu is the selection of items you have chosen to show. The site structure is the map of every page on the site, including those that can't be reached from the menu. On an e-commerce site with a few thousand product pages the two diverge straight away.
Google doesn't present it that way. It says it uses the number of links to follow to reach a page and the number of links pointing to it to infer its importance relative to the rest of the site. It affects discovery and priority, not a score given to the structure.
If you don't need them indexed, yes. Google recommends preventing crawling of filtered URLs and leaving only the unfiltered category page and the individual product pages crawlable. If some combinations need to rank, keep them few, with a standard separator between parameters and the filters always in the same order.
