Site architecture for SEO: every page gets one job.

Site architecture is how a website’s pages are grouped and addressed, and how they link to each other, so that every page has one job and both visitors and Google can find it. It’s the least visible part of organic SEO.

One search, one page

When two or three pages target the same search, Google has to pick, and it may switch between them or rank each lower than a single page would. You can see it happening: the address Google shows for a search changes from week to week, or two of your pages collect impressions for the same search. I usually fix it by merging the competing pages into the strongest one and redirecting the old addresses, so links and visitors follow.

Each group of searches should have exactly one page competing for it. The list of which search belongs to which page comes from keyword research.

Grouping pages the way customers think

I group pages by service. Take a plumbing company. The shape might be:

• Home links to each service group.
• Water heaters (a group page) links to repair, tank replacement and tankless installation.
• Each service page links back to its group and across to closely related services.

Pages linked from the home page and the main menu are the ones Google reaches most readily and tends to treat as most important. A service page that takes five clicks to reach is, in effect, marked as minor.

Internal links and their wording

One page vouches for another by linking to it, and the words in the link tell Google what the target is about. “Tankless water heater installation” as link text says more than “click here”. I add links inside the text of related pages, where a reader would want them, in addition to the menu and footer.

I also look for orphan pages, which are pages nothing else on the site links to. Google has little to go on in finding or valuing them, and they tend not to rank. Linking is separate from indexing and crawling, which is whether Google can fetch and store a page at all.

Addresses that stay put

A page’s address (its URL) should be short and readable, and it shouldn’t change. “/water-heaters/tankless-installation/” is better than “/services-2/page-id-417/”. Tidiness alone is a poor reason to rename one, though. Every change means a redirect and a period while Google reads the page again, so I leave a working address alone unless the structure around it’s being fixed anyway. I plan a large restructure like a site migration.

Service and city pages without the copies

The quickest way to wreck a structure is to multiply every service by every city. Twelve services across twenty cities gives 240 pages that differ by one word. Google’s spam policies describe pages like that, built to catch variations of one search, as doorway pages. A city page earns its place when you have something true and specific to say about working there, and that test decides which service area pages I build.

How I test structural changes

I make small changes one at a time. Adding three or four internal links to a page that sits low is a clean test, and I check its position about two weeks later.

Merges are bigger, and some merged pages need longer than that to settle, so I’ll say when a result is still unclear. Before a merge, I record the positions and clicks of every page involved, so that afterward I can judge it on numbers.