- Map a subject as a set of distinct search intents before you write, and merge any two queries that one page would fully satisfy.
- Publish three or four supporting pages first, then the pillar once the pages it links to are live, so no link points at a gap.
- Audit a cluster's internal links monthly and fix any relevant page with fewer than two inbound links from inside the cluster.
Ask three SEO consultants what topical authority is and you will get three answers, and one of them will tell you Google scores it. Google doesn’t. It has never published a topical authority metric, and its documentation doesn’t name one. The pattern the term describes is real, though, and easy to observe: a site that covers a subject thoroughly tends to rank for the individual questions inside that subject, and its new pages start earning impressions faster than the same pages would on a site with no track record there. Topical authority is the name for that compounding effect. The useful question is not how to get a score. It’s what causes the effect, and how to produce it on purpose.
The short version: topical authority compounds through three things. You need a coherent map of the subject, enough depth on each part of it, and internal links that show how the pieces relate. When those line up, every new article makes the others work harder. When they don’t, you end up with forty posts that each rank for almost nothing.
What topical authority is, and what it isn’t
Practitioners use the phrase for a site that reads as a reliable source on one subject. “Reads as” is doing real work there. A few observable signals sit behind it:
- Consistent vocabulary. A page on reconciling bank feeds that uses the same terms as your other pages on accounts receivable (payment matching, aging reports, write-offs) belongs to a coherent body of work. A page on the same subject that wanders into unrelated jargon does not.
- Coverage of the follow-up questions. A reader who finishes one page and immediately has the next question is a sign the subject has been mapped. Sites that answer only the headline query leave that reader to go elsewhere.
- Clear structure. How pages link to each other, and which page the site points to as the main answer for the subject, tells search systems where the center of the topic sits.
None of those is labeled “authority,” and none of them needs a number. They are the parts of the effect you can actually influence.
Topical authority is not a license to write about everything, and it is not a shortcut around expertise. Google’s guidance on creating helpful content asks whether a page shows first-hand expertise and a real depth of knowledge, and whether the site has a clear focus. You cannot write deep coverage of a subject nobody on your team understands. Depth shows up in specifics: the edge case, the failure mode, the number someone actually checked. Templates don’t produce those.
Breadth is not authority: a site with 200 articles across 30 unrelated subjects has topical authority on none of them. It has 30 thin, weak sites sharing one domain.
Why depth compounds: the mechanism
It helps to know what is actually happening, because the mechanism tells you what to build. There are three parts.
Shared relevance. When ten pages on one subject use consistent terms and link to each other, they corroborate one another’s relevance. A single strong page with nothing around it has no corroboration. A dense cluster does. In practice this often shows up as new pages in an established cluster being indexed and placed sooner than an isolated page would be. “Often” is the honest word here, because it varies from site to site.
Link flow. Internal links move context and, through it, some of the ranking value a page has accumulated. If a supporting article picks up a few external links, a well-placed internal link from it can help the pillar it points to. Links also give crawlers a route to each page. A page with no internal links depends on the sitemap alone, which is a weak position for anything you want to rank.
Session depth. A reader who lands on one answer and finds the next one nearby stays longer and comes back. You can see this in analytics as pages per session from organic entry points. It is the most reader-facing part of the effect, and the one most often forgotten when teams plan a cluster.
Take two freelance-invoicing sites, each publishing twelve articles on the subject over six months. Site A writes posts as ideas occur, each linking only to the homepage. Site B picks the same twelve topics, orders them deliberately, and links every post to a hub page and one or two siblings. Both have the same writing quality. Site B’s posts get indexed faster, pick up more long-tail queries, and show more pages per session. This is an illustration, not a benchmark, but it is the pattern the mechanism predicts. The difference is the wiring, not the prose.
Map the subject before you write anything
Most clusters fail at the map, not the writing. A team starts with a keyword list, publishes against it one item at a time, and discovers six months later that three of the pages were the same page. Mapping prevents that before any article exists. This is the process I would follow.
-
Write the boundary in one sentence. For example: “This cluster covers how freelancers invoice clients, chase late payments and track what they’re owed. It excludes tax filing, payroll and software comparisons.” The boundary tells you what to cut. Without one, every adjacent topic looks like it belongs, and the map sprawls into the detour problem described later.
-
Collect the questions people actually ask. Four sources, roughly in order of value: - Search Console queries for pages you already have. Pay attention to queries where you sit at positions eight to thirty. They show demand you are close to capturing. - Autocomplete and “People also ask” results for the head term, run several times with different modifiers. - Support tickets, sales call notes and onboarding questions. This is the most underused source, because customers describe problems in their own words, and those words are often the ones searchers type. - Competitor sitemaps, read for coverage gaps. Use them to see which sub-questions nobody answers well, not to copy their outline.
-
Group by intent, not by keyword. Two queries belong on one page when one page can satisfy both searchers. The test is quick: search both and compare. If the top results are largely the same URLs in the same format, it is one page. If one result is a definition, another a calculator and a third a comparison, you have two intents and two pages. “Invoice template” and “free invoice template download” share an intent. “How to chase an unpaid invoice” and “can I charge interest on an unpaid invoice” do not, and the second usually needs a jurisdiction-specific caveat the first does not.
-
Assign each page a role. Every page in the cluster needs a job. The pillar’s structure gets its own treatment in our guide to pillar pages and content clusters, but the roles below are the ones I use when drafting a map.
| Role | Job on the page | Typical scope | Links out to |
|---|---|---|---|
| Pillar | Maps the whole subject and names each sub-question | Broad; covers every sub-topic in a few paragraphs | Every supporting page |
| Core supporting | Fully answers one search intent | Narrow and complete | The pillar and one or two siblings |
| Utility | Gives the reader a tool, template or checklist | Narrow; the artifact is the value | The pillar and the explainer it depends on |
| Decision | Helps a reader choose between options | Narrow and opinionated | The pillar and the pages it compares |
- Rank the candidates by value. For each planned page, estimate three things: its demand (Search Console impressions for related queries, or a research tool where you have no data), its winnability (are the current top results forums and outdated posts, or specialists with strong pages?), and how well a reader who finds it fits what you offer. Multiply the first two and weigh the third. Tools such as AutoSEO Stack’s Google research pull live query data to speed this up, but a spreadsheet and an afternoon in Search Console will give you a usable first map.
A worked example: one cluster from map to links
Say the subject is unpaid invoices, and the site sells invoicing software. The map might look like this.
- Pillar: how to handle unpaid invoices, from the first reminder to write-off. It covers each stage briefly and links to every supporting page.
- Core supporting: net 15 versus net 30 payment terms and what each does to cash flow; how to read an invoice aging report; when to write off a debt and how to record it.
- Utility: a payment reminder sequence, three templates with the tone shifting at each stage; a late-fee calculator.
- Decision: should you charge a late fee at all? The factors that matter, and the cases where a fee costs you a good long-term client.
- Specialist, with a caveat: small claims for unpaid invoices. The rules vary by jurisdiction, so this page should say which parts change by location and link to the authority that sets them rather than pretending there is one answer.
Build the calculator early, since other sites link to tools more often than to prose. Build the specialist page last, once the cluster’s links can support it.
Sequence the publishing so every link lands somewhere
Most guides tell you to publish the pillar first. For a new site I would reverse that, and say so plainly. A pillar published before its supporting pages links to pages that don’t exist yet, so you either leave dead links or strip them out later. A pillar published after its supporting pages has something to link to from day one, and the supporting pages can link up to it as soon as it goes live.
The exception is a pillar that a sales conversation or a paid campaign needs right now. In that case publish it first and accept the gaps. The sequence I would use otherwise:
- Publish three or four supporting pages, starting with the best winnability scores. Choose the narrow intents where current results are weak. These are your fastest signals.
- Read Search Console for two to four weeks. Note every query getting impressions, including the ones you didn’t target. Some will be new pages on the map. Others will tell you an existing page should be split or merged.
- Publish the pillar once most supporting pages are indexed. Then add a link to it from every supporting page already live. This is an edit to older pages, and it is the step teams forget to schedule.
- Publish the remaining supporting pages in one tight batch. Each one links up to the pillar and across to one or two siblings. A batch makes the link graph complete at once. Spreading the same pages across five months leaves the cluster looking half-built for most of that time, and readers and crawlers both see the gaps.
- Link forward from the early pages to the late ones. The first pages are your strongest link sources because they have been around longest. Using them to point at the newer pages is the cheapest boost in the whole process.
- Measure before you start the next cluster. Starting a second cluster before the first one has been read is how sites end up with three half-finished maps.
On timing, expect a new site to need months before a cluster shows clear returns. The first returns are usually impressions on supporting queries, well before the pillar reaches the top ten for its head term. I would not promise anyone a date.
Internal links: the wiring that makes depth count
Internal links do three jobs. They give crawlers a path to every page. They tell a reader the next step. And they signal which page is the main answer for a given query. Most sites do the first job and ignore the other two. Anchor text and placement mechanics get their own treatment in our internal linking strategy, so here I will stick to the rules that matter specifically for clusters.
- Link where the reader needs the next step. A link earns its place when a reader would plausibly click it mid-sentence. A paragraph about payment terms that links to the payment-terms page passes that test. A sidebar of twenty related links does not.
- Make the pillar the hub in both directions. Every supporting page links up to the pillar, and the pillar links down to every supporting page. This two-way structure is what tells a crawler which page is the center of the topic.
- Use descriptive, varied anchors. “Our guide to net-30 payment terms” is a good anchor. “Click here” says nothing. Repeating one exact-match phrase as the anchor on every page reads as engineered to any human who notices it.
- Ignore the generic links when counting. A link from the footer or a “latest posts” widget appears identically on every page, so it carries almost no topical signal. Count links from pages that are actually about the subject.
- Hunt orphans. Crawl the site and filter for pages with zero inbound internal links. Those pages depend entirely on the sitemap, and they are often the pages that never rank.
Run the orphan check monthly: any supporting page with fewer than two inbound links from relevant pages in its own cluster is a candidate for linking work or merging. Links decay as you add and retire pages, so a launch-day check is not enough.
Where topical authority breaks
Four failure modes come up often enough to name.
Spoke pages written to fill a box
A map with a box for every sub-question invites you to fill every box. A box with no real intent behind it becomes a page with no reason to exist. Nine complete pages beat fifteen where six are thin. Pages written only to complete a diagram look a lot like the pattern Google’s scaled content guidance describes, and they dilute the strong pages around them.
Cannibalization
Two pages with the same intent compete for one query, and often neither wins. Search Console shows this as a single query alternating between two URLs over time. The fix is to merge the pages, keeping the stronger one and redirecting the other, or to sharpen one of them toward a different intent. Rewriting both with new synonyms rarely helps, because the intent has not changed.
Detours
A freelance-invoicing cluster that drifts into productivity apps and retirement planning might each bring defensible traffic, and still blur what the site is an authority on. Set the boundary once and hold it. Treat an off-boundary topic as a decision about a new cluster, not as a one-off post that slips in because it ranks.
Subjects with rules that go stale
Tax thresholds, platform policies and legal rules change. A cluster built on them decays faster than most, and one outdated rule makes the whole cluster look abandoned. Give every rule-based page a review date in your tracker, and update the pillar first, because it carries the most links and a stale pillar drags the rest down with it.
How to tell whether it’s working
There is no topical authority score to check, so measure the observable parts of the effect.
- Cluster query coverage. Count the distinct queries where any page in the cluster gets impressions. A rising count means search systems are starting to associate your pages with the subject.
- Top-ten rate across cluster queries. Track the share of cluster queries where you rank in the top ten, rather than any single position. One page’s exact position moves around too much to read.
- Time to first impression for new cluster pages. Compare it with how long your earlier, unrelated posts took. If clustered pages are seen sooner, the shared-relevance effect is probably doing its job.
- Pages per session from organic entry. Check how many pages visitors view after landing in the cluster. This is the reader-side measure, and it tells you whether the internal links do anything for people.
- Unwritten map entries. Each planned page that isn’t written is a visible gap, and filling it is a measurable change.
Give a cluster a fair read after a full cycle: all supporting pages indexed, the pillar live, links in both directions, and at least two months of data. Reading it after two weeks tells you about indexing, not the cluster.
Where this leaves you
Topical authority is the result of a cluster built well. It is not a setting, a plugin or a word count. Build from a bounded map, give each page its own intent and role, link on purpose in both directions, publish in an order that keeps every link pointing at something real, and measure coverage instead of chasing a score.
That is the discipline AutoSEO Stack is built around. It researches what buyers actually search, writes each page to a per-page quality bar in a human voice, and publishes on a schedule, so a cluster gets built one honest page at a time instead of as a pile of posts. If you want to see what that looks like on your own site, start with one subject, one map and a look at the features. Then judge the output the way this article suggests: would each page be worth reading on its own?