Topical authority: the site architecture behind 500 URLs

Strip the head term; the modifier decides the folder — and the register decides whether a page exists at all.

You cannot look up your topical authority. There is no score, no dashboard and no API for it.

What you can do is build the structure that makes the description true. That is an architecture job, and it is the part almost nobody writes about — because writing about it honestly means publishing your own folder tree.

So that is what this post does. It sits in the architecture half of our SEO guide, and the worked example throughout is this website: roughly 500 published URLs, deliberately built hub-and-spoke, with the rules written down before the pages existed.

What topical authority is — and what Google has actually said

Topical authority is a shorthand SEOs use for a pattern they observe: sites that cover one subject exhaustively tend to rank for more of that subject's long tail than sites that cover it once.

That is a description, not a mechanism. Here is the difference between what gets claimed and what Google documents.

Read together, the honest position is this: Google documents that a site earns signals from its own content and that having a focus is a quality marker. It has never published a topical-authority metric. Build for the first two and ignore anyone selling you a score for the third.

Topical authority vs domain authority

These get used as synonyms and they are not even the same kind of thing.

• Domain Authority is a third-party tool score. It is Moz's 0–100 prediction of ranking strength, calculated mostly from link data. Google does not produce it, receive it or read it. Ahrefs, Semrush and SE Ranking each publish their own competing version.
• Topical authority is subject-scoped, not site-scoped. A site can be the most complete source in India on NBFC marketing and have nothing useful on mattresses. One tool score for the whole domain cannot express that.
• Google has pushed back on site-level authority language for years. "In general, Google doesn't evaluate a site's authority," John Mueller said in a 2019 hangout (Search Engine Journal, 13 Jun 2019). Treat all of these scores as competitive-research shorthand, never as a target.

Useful rule: if a number can be bought with links alone, it is not describing whether you answer a subject completely.

Seven folders and no more

This is this site's own build, not a template. Every URL on digitalhangover.in lives in one of exactly seven folders:

Capping the namespace is the whole point. Seven folders means every new page has exactly one defensible home, and "where does this go?" becomes a ten-second decision instead of an argument. Architectures sprawl when the eighth folder gets created for one page that did not fit — and then nothing ever fits again.

One entity, one owner URL, forever

An entity is a thing people search for — a concept, a method, a tool, a place. The rule is one sentence: each entity gets one owning URL, and no second page may be built for it. We keep that written down as a cannibalisation register, checked before any page is created.

The register is cheap to maintain and it does the thing no audit tool does: it stops the duplicate before it is written. If you already have overlapping pages, that is a different job — see how to find and fix keyword cannibalisation. The register is the prevention, not the cure.

The modifier decides the folder

Here is the routing rule we actually use. Strip the head term out of the query. Whatever words are left tell you which folder the page belongs in.

That last row is a rule with teeth: if the answer can be computed, build the tool instead of the article. Someone searching "cpc calculator" wants a box to type into, not 1,400 words about the formula. That one decision is why this site has 20+ live calculators in the tools library rather than twenty more posts.

If your routing rule starts generating hundreds of near-identical pages from a template, you have crossed into programmatic SEO, which has its own quality rules. The modifier test is for human-written pages.

The link rule — and the one we broke

Two rules, both absolute. A pillar must link down to 100% of its children. Every child must link back up to its pillar. If both edges do not exist in the HTML, the cluster does not exist.

Google is explicit about how it finds pages: they "are discovered when Google extracts a link from a known page to a new page" (Google Search Central, last updated 18 Dec 2025). A sitemap entry gets a URL crawled. It does not tell anything that the page belongs to a group.

Now the honest part. When we audited this site in August 2026, none of the then-45 live blog posts linked up to a pillar. The clusters were correct in the plan and absent from the markup. We had built the downward links, published the posts, and never gone back for the return edge.

That gap had to be fixed page by page afterwards, and it is the single most common way a topic cluster fails in practice. The plan looks like a cluster. The crawl sees a pile. Two related reads: what makes a page an orphan, and the mechanics of fixing it in our internal linking guide. On large sites the same gap wastes crawling, which is covered in crawl budget.

One more rule from the same family: a URL is permanent. When a page needs updating, we refresh it in place and never republish it at a new slug. Google's own URL guidance only tells you to keep structure simple (Google Search Central, last updated 10 Dec 2025) — it does not tell you to freeze slugs. That rule is ours, and the reason is link equity: every slug change adds a redirect hop and risks the external links pointing at the old address.

How to plan your first cluster in five steps

1. Name the entity, not the keyword. "Topical authority" is an entity. "How to build topical authority fast in 2026" is a phrasing of it. Entity SEO explains why that distinction changes what you build.
2. List every question in the subject. Pull them from keyword research, autocomplete, people-also-ask and your own sales calls. Long-tail queries are where completeness is actually won.
3. Route each question with the modifier rule. Guide, blog, research or tool. Do it for the whole list before writing anything.
4. Check the register. If an entity already has an owner URL, the new question becomes a section inside that page — or it does not get made.
5. Wire both directions before publishing. Pillar links down to the child, child links up to the pillar, same deployment. Content mapping is where you record which URL owns what.

The content-side version of this planning — angles, briefs, what each page actually says — lives in our topic clusters guide. How the cluster fits a wider plan sits in SEO strategy and content marketing strategy. This post is only the skeleton.

Expect the usual organic timeline, not a switch: meaningful movement in roughly three to six months, compounding after that as the cluster fills in.

What a capped architecture gives you next

The payoff is not a score. It is that decisions get cheap.

A new question arrives, the modifier rule places it in seconds, the register says whether it already has an owner, and the two links get written in the same commit. Refreshes replace rewrites, because the URL never moves. And when one URL owns one question, there is only one version of your answer for anyone to find — a reader, a crawler, or an AI assistant quoting you — rather than four near-duplicates competing to be the one.

We are not going to show you a traffic chart for that. We have not measured the architecture against a counterfactual version of this site and we are not going to pretend otherwise. What we can show you is the register, the folder list and the audit that caught our own missing links — which is more than a score would ever tell you. Start with how search engines read meaning, then write your register before your next page.