Faragiri

Learn Site structure SEO · Topic clusters, URLs

Of 612 pages on a Cornish kayaking site, 477 were made by the software. What the links said the site was about was not what anybody had decided.

2Branches
2Nodes below
3Depth

Cadgwith Charts publishes sea kayaking route guides for the Cornish coast and had 612 pages in March 2026. 47 were route guides that earned clicks. 88 were tide and access notes. The other 477 were made by the software: 194 tag archives, 141 paginated list pages and 142 duplicated print versions. Only 135 of the 612 pages were written by a person.

A site is not what it publishes, it is what it links. A site's structure is the set of links, not the set of folders, and on that measure Cadgwith was a site about tag archives with some kayaking on it.

One hub, and everything under it pointing up

The shape that works is unglamorous. One hub per subject, and every page under it links up to that hub by name. The Lizard hub sits at a routes directory and 41 route pages point up at it using the words a person would use for it. The north coast hub has 23. Each route page also links sideways to the two or three routes somebody would actually paddle next, which is a judgement about tides rather than about search.

Hub and spoke shape

The shape a 612 page site actually needed

Two hubs, the notes they call on, and nothing else at the top level. Every route page points up at its hub by name and sideways to adjacent paddles.

cadgwithcharts.co.uk

  • /routes/The only directory anything outside the site links to
    • /routes/lizard-peninsula/The hub. 41 route pages point up at it by name
      • /routes/lizard-peninsula/cadgwith-to-kennack/
      • /routes/lizard-peninsula/mullion-to-porthleven/
    • /routes/north-coast/The second hub, with 23 routes under it
  • /tides/88 access notes, each linked from the routes that need it
  • /tag/194 archives, generated by the theme and linked from every page

Forty-one routes linking up with the same hub subject concentrates description so a small site can out-rank a sprawling one.

What that arrangement does is concentrate description. Forty-one pages saying Lizard Peninsula sea kayaking routes in a link is forty-one sentences about what the hub is, and no page anywhere had to claim it about itself. A site with no hubs has every page making its own case alone, which is why a large site can be comprehensively out-ranked by a small one that has decided what it is.

The 194 that had to go

Tag archives are the standard way a site accidentally competes with itself. 194 tag archives were 194 pages competing with the hub they were made from: a tag page for lizard listing eleven routes, next to a hub about the Lizard listing forty-one, and nothing on the site to tell an engine which of the two was the real one.

March 2026

What the 612 pages actually were, before the 194 tag archives went

Drawn as one bar because the 194 generated tag archives outnumbered the real guides and competed directly with the hubs.

Pages on cadgwithcharts.co.uk, March 2026.

  1. Tag archivesGenerated, linked from every page, competing with the hubs194
  2. Duplicated print versionsA second address for every guide, saying the same thing142
  3. Paginated list pagesThe only route to a third of the guides until the hubs were fixed141
  4. Tide and access notesWritten by a person, and linked from the routes that need them88
  5. Route guidesThe pages anybody came for47

Deleting the 194 tag pages in February 2026 moved the Lizard hub quietly from position 9.4 to 6.8 by May.

They went in February 2026, and the honest account of what happened is that nothing dramatic did. The Lizard hub moved from position 9.4 to 6.8 between February and May 2026, and the eleven-route tag page stopped appearing at all. Removing pages that were never read is a quiet change with a slow payoff, which is why it keeps not getting done.

Shallow is about clicks, not slashes

The advice to keep URLs shallow gets read as an instruction about punctuation. It is not. Shallow means few clicks from the home page, not few slashes, and a page at a four-segment address linked from the main menu is shallower in every sense that matters than a two-segment page reachable only from page nine of a list.

Clicks, not slashes

Why 141 paginated list pages buried a third of the guides

Shallow means few clicks from the home page, not few URL slashes; forcing crawlers through nine list pages hides routes once rel next/prev retired in 2019.

Changing a directory name without a redirect throws away what the old URL earned.

Which makes pagination a structural problem rather than a cosmetic one. Cadgwith's 141 list pages were the only route to a third of the guides, and rel next and prev stopped being used in 2019, so nothing about the markup was telling anybody those pages belonged together. The fix was not markup. It was that each hub now links directly to every route under it, so nothing depends on a crawler walking to page nine.

The other half of a structural decision is that addresses have to survive it. A URL that changes without a redirect loses what the old one earned, and the tidying instinct that renames a directory to something nicer is the most expensive twenty minutes available on a site this size.

Where this stalls: the menu is the only structure anybody edits

The stall is that navigation is visible and internal linking is not. A menu gets argued over in a meeting; the links inside a paragraph get written by whoever wrote the paragraph, on a Tuesday, with no view of the site. So the declared structure and the actual structure drift apart, and the actual one is the one that counts because it is the one made of links.

Declared vs built

Teams argue over top menus in meetings, while footer tag links and ad hoc paragraph links quietly outweigh them.

  • Main navigation menuSays routes, tides, about; debated in meetings
  • Intended folder hierarchyCleanly organised in the editorial plan
What teams inspect
  • Footer tag links on every pageGenerated by the theme, making tags the most linked pages
  • In-paragraph linksWritten one Tuesday at a time with no whole-site view
  • Actual site structureDefined by every internal link, not by the menu alone

Nobody was wrong in any meeting; the theme simply linked tag archives from the bottom of every page.

At Cadgwith the menu said routes, tides, about. The links said tags. Nobody was wrong in any meeting; the archives were simply generated by the theme and linked from the bottom of every page, which made them the most linked pages on the site.

Seeing the shape rather than the sitemap

Ahrefs Webmaster Tools crawls a site you have verified and reports crawl depth and the pages nothing links to, which is the structure as built rather than as intended. It is free for a verified site, and it replaces paying for a full crawl of something you already own.

Seeing the actual shape

Two free crawl checks and the subject decision neither tool can make

Ahrefs Webmaster Tools shows depth and orphans, Search Console shows submitted vs found URLs, and editors decide what counts as a hub.

  • Ahrefs Webmaster Tools
  • Search Console indexing
  • Subject boundary call
  • Ahrefs Webmaster ToolsFree for verified sites; maps real crawl depth and orphan pages
  • Search Console indexingExposes gaps between sitemap URLs and generated pages found
  • Subject boundary callHuman judgement on whether the Lizard is one paddling hub or three

A large gap between submitted and found pages almost always means the software is generating URLs nobody meant to publish.

Google Search Console answers the other half. Its indexing report separates what was submitted from what was found, and a large gap between the two is usually a site generating pages it never meant to publish. It is free, and it replaces counting pages by hand and believing the total.

Neither tool can say which page should be the hub. That is a decision about the subject, and at Cadgwith it took an argument about whether the Lizard is one paddling area or three.

The five structural decisions

The hub and its forty-one is a hub, its supporting pages, and the links between. The addresses those pages sit at are readable, stable and shallow, with the redirect rule attached.

The five structural decisions

Five layers of site structure, resting on redirects and canonicals

From surviving URL changes to pruning 141 lists and 194 tags, each layer depends on changing one thing at a time if you want to measure it.

  1. Menus and anchor textClear top hierarchy plus page-level anchor phrasingVisible signal
  2. Pruned pagination and tagsRemoving 141 list splits and 194 competing tag archivesSignal hygiene
  3. Readable, shallow URLsStable paths kept within a few clicks of homeAddressing
  4. Hub and spoke linksForty-one routes reinforcing one central hubConcentration
  5. Redirects and canonicalsDecides whether any restructure survives without dropping 2.6 placesFoundation

Cadgwith changed three other things in the same quarter the Lizard hub moved 2.6 places, making single-cause attribution impossible.

The visible half is menus as the clearest statement of your hierarchy, and the 141 list pages are long lists, without splitting your own signals. The 194 are pages that dilute the ones that matter.

The anchor text carrying all of this is decided one page at a time, at what a single page controls about its own ranking, and the redirects and canonicals that make a restructure survivable are the layer that decides whether any of the rest counts.

The Lizard hub moved 2.6 places in the three months after 194 pages were deleted, and Cadgwith changed three other things in the same quarter. Which of the four moved it is not knowable now, and the only way to find out would have been to do them one at a time and wait.

Inside this, 2

Alongside this

Above this