Content architecture beats content volume

When a site isn't getting the search traffic it wants, the most common instinct is to publish more. More blog posts. More landing pages. A content calendar with a post every day. With AI writing tools, that instinct has gotten cheaper to act on, and more dangerous.

I've come to believe most sites don't have a volume problem. They have an architecture problem. The pages they already have are competing with each other, poorly connected, and unclear about what each one is for. Adding more pages to that structure mostly adds more confusion.

Here's how I think about content architecture, and the practical steps I use to fix it.

Every page needs a job

Before anything else, I want to know what each page is for. Not its topic. Its job.

A page's job is a combination of three things:

• The question or intent it answers. One primary intent, not five.
• Who it's for, and where they are. Someone learning the basics, someone comparing options, someone ready to act.
• What should happen next. Read a related page, start a conversation, fill out a form, sign up.

If you can't state a page's job in one sentence, that's a finding. It usually means the page is trying to do too much, or it's a near-duplicate of another page, or it exists because someone thought the site needed more content.

A simple exercise: export every indexable URL into a spreadsheet and add a column called "job." You'll find the problems quickly.

Pillars and clusters

The structure I keep coming back to is the pillar-and-cluster model. It's not new, and it's not magic. It works because it mirrors how people actually learn about a topic.

• A pillar page covers a broad topic comprehensively. It's the best single page on your site for that subject, and it gives a clear overview with enough depth to be useful on its own.
• Cluster pages go deep on specific subtopics or questions. Each one answers one thing thoroughly.
• Links connect them deliberately. The pillar links out to each cluster page where relevant. Each cluster page links back to the pillar and to closely related siblings.

The result is a structure that makes sense to a reader, and a clear signal to search engines about which page is your main resource on the topic and how the supporting pages relate to it.

This is also how this site is built. Posts are grouped around a few themes, with a pillar for each. The one on getting cited in AI answers is the pillar for search and AI visibility, and the posts that followed link back to it.

Internal linking is architecture

Internal links do more work than most people give them credit for. They help search engines discover pages, understand how pages relate, and judge which pages matter most. They also guide readers to the next useful thing.

What I look for:

• Orphan pages. Pages with no internal links pointing to them. If you don't link to it, you're telling everyone it doesn't matter.
• Descriptive anchor text. "Our guide to filing deadlines" tells both readers and search engines what's on the other side. "Click here" doesn't.
• Links in the body, not just navigation. Contextual links inside the content carry meaning that a sitewide footer link doesn't.
• Important pages getting important links. Your key pages should be reachable within a few clicks from the homepage and linked from relevant content.
• Broken and redirected internal links. Link directly to the final URL. Redirect chains waste crawl effort and slow users down.

A crawl with any decent SEO crawler will show you inlinks per page. Sort ascending. The bottom of that list is usually where the forgotten pages live.

Cannibalization: when your pages compete with each other

Keyword cannibalization is what happens when multiple pages on your site target the same intent. Search engines have to choose between them, and often they choose inconsistently. You end up with two mediocre pages instead of one strong one.

How to spot it:

• In Search Console, filter the Performance report by a query and look at the Pages tab. If several URLs share impressions for the same query, and the ranking URL keeps switching, that's a signal.
• In your content list, look for titles that are variations of each other. "How long does the process take," "Process timeline explained," and "What to expect: timeline" are probably one page.
• In your own navigation, if you can't tell which page to send someone to for a question, neither can Google.

Not every overlap is a problem. Pages can share keywords while serving clearly different intents. The question is whether they're competing for the same job.

Pruning and merging

The most effective content work I've done has often been removing and combining, not adding.

For each weak or overlapping page, there are a few options:

Situation	What to do
Two pages, same intent, both partial	Merge into one strong page, 301 redirect the other
Outdated but still relevant topic	Update it properly and refresh the date
No traffic, no links, no clear job	Remove it (410 or 404), or noindex if it must stay for users
Thin page with a real intent behind it	Expand it until it's actually the best answer
Useful for users but not for search	Keep it, noindex it, keep linking to it where helpful

When merging, move the best material from each page into the survivor, update internal links to point directly to it, and redirect the retired URL. Don't redirect everything to the homepage. Send each old URL to the closest genuine equivalent.

Pruning feels risky because it looks like losing pages. In practice, a smaller site where every page has a job is easier to crawl, easier to maintain, and easier for both people and search engines to understand.

Why this matters more with AI

AI answer engines pull passages, not whole sites. A clean architecture helps here in a few ways. A single authoritative page per topic is more likely to hold the clearest passage than five overlapping ones. Clear internal structure helps crawlers find and understand your best material. And a focused site makes it easier to understand what you're actually an authority on.

The flip side: it's never been easier to generate hundreds of pages. Google's spam policies call out scaled content produced primarily to manipulate rankings, regardless of how it's created. Volume for its own sake is a liability, not an asset.

How I'd start

1. Export every indexable URL and give each one a job. Flag the ones you can't.
2. Group pages by topic and pick one pillar for each group.
3. Find cannibalization in Search Console and decide what merges.
4. Prune or noindex pages with no job.
5. Rebuild internal links: pillars to clusters, clusters to pillars, fix orphans and redirects.
6. Only then plan new content, to fill real gaps in the structure.

New content is worth writing when it fills a gap in a structure that makes sense. Until then, the best page you can publish is often the one you merge.