Topical Authority Without the Content Farm: How We Plan a Cluster

Topical Authority Without the Content Farm: How We Plan a Cluster

TL;DR: Topical authority is usually explained as coverage: publish every subtopic and the hub gets stronger. That version produced the library we deleted a quarter of in August 2026. What replaced it is a two-column planning sheet where every planned article carries both a validated search volume with its pull date and a named first-hand data source. If either column is empty the article is cut. Planning this cluster that way cut 20 candidate topics to 13, and the cuts were all on the volume side in Chinese and the data side in English.

Cluster planning fails in a predictable way. Somebody exports a keyword list, groups it by parent topic, and the group becomes the content calendar. Nothing in that process asks whether you have anything to say.

The two columns

So we added a second question to the sheet, and it turned out to be the one that does all the filtering.

Our planning sheet has one row per candidate article and two columns that both have to be filled before anything gets written. Column one is the target term with its search volume, keyword difficulty and the date the numbers were pulled. Column two names the specific first-hand data the article will carry: a client account, a measurement we ran, a support ticket, a bug we caused. Neither column is negotiable, and the second one is the harder gate. Coverage is easy to fake and data is not.

The reason for column two is the same reason coverage-first planning stops working. If every fact on the page came from public sources, a reader is better served going to those sources, and search engines have been moving visibility toward destinations and away from intermediaries. Google's guidance on creating helpful content↗ frames this as the "who, how and why" test, and the practical version is blunter: remove your citations and see whether anything is left.

What the two columns did to this cluster

Planning the cluster this article belongs to, we started with 20 candidate topics across two languages and finished with 13, a cut of 35%. In Traditional Chinese, every gambling-specific SEO↗ term we tested returned effectively no demand: an Ahrefs pull on 2026-08-27 for the Taiwan market returned 50 monthly searches for one spelling of "casino SEO" and 0 for six other variants. In English, the gambling-qualified long tail was real but thin, with terms like "igaming seo audit" at 60 and "casino seo strategy" at 10 monthly searches.

The volume that existed sat somewhere neither list expected: generic Traditional Chinese SEO terms with very low difficulty. "sitemap" returned 1,700 monthly searches at difficulty 14, "robots.txt" 500 at difficulty 3, "structured data" 300 at difficulty 5. Those numbers also carried real click counts, which matters because impressions in our English data are heavily agent-driven.

That result inverted our prior. We expected the vertical-qualified terms to be the defensible ones because they match what we sell, and the data said the defensible terms were generic method terms where the difficulty scores were near zero. The vertical stays in the writing, as the source of examples and first-hand data, rather than in the target term. That distinction is worth stating precisely, because it is the difference between an article that a gambling operator reads and forwards to their agency and an article that only ranks for a term nobody in the industry searches. The first kind earns links and citations from people who have budget. The second kind sits at position 4 with 134 impressions and one click, which is exactly what our best-performing English page did over the 90 days to 2026-08-26 before we understood the query mix behind it.

Hub and spoke, with the hub already built

The order most guides recommend here is exactly backwards for anyone who has been publishing for a year or more.

The usual advice is to write the pillar first. We did the opposite here, because a pillar already existed and was ranking at position 22 for the head term. Retrofitting it cost a fraction of writing a new one and avoided splitting the same query between two of our own pages.

The retrofit itself was instructive. Running our acceptance script against the existing English cluster on 2026-08-27 returned 2 passes out of 20 files, and 11 of the failures were the same single check: fewer than three distinct external authority domains cited. Adding real sources to those 11 files took under an hour and moved the pass rate to 8 out of 20. Writing a new article would have taken a day and moved nothing.

That is the part of topical authority that nobody sells: the cheapest authority gain available to most sites is repairing what is already indexed. In our case one hour of edits moved 55% of the failing files across the line, against roughly 25% of a working week for a single new article.

Internal links are the actual mechanism

A cluster is a link graph, not a list. Skipping the sideways links is the single most common omission we find in inherited content plans. Each spoke links up to the hub, sideways to two siblings, and once to a commercial page. The sideways links are the ones teams skip, and they are what turns a list of articles into a structure a crawler can read as a topic.

Two rules keep it honest. Links carry the site canonical form, including the trailing slash, because a link in the wrong form takes a redirect and hides genuine 404s. And no target gets linked more than three times from one article, because past that it stops being a signal and starts looking like a footer.

None of this is complicated. It is just unpopular, because the sheet says no far more often than it says yes.

FAQ

How many articles does a cluster need? Fewer than you think. The number is set by how many you can back with data. We planned 13 across two languages because that was how many had a named first-hand source, not because 13 is a threshold. Our own top 9.4% of pages carry 93.5% of clicks, so depth beats breadth by a wide margin.

Is topical authority a real ranking factor? Not as a named signal. It is a description of an outcome: sites that cover a topic with depth and get engagement tend to rank across it. Search Engine Land↗ is a reasonable place to track how the discussion has moved.

Should I write the pillar first? Only if you do not already have one. If a page is ranking on the head term, retrofit it. Publishing a second page for the same query splits your own results.

Retrofit the ranking page first, every time. A new URL competing with your own is the most expensive mistake in cluster work.

How do I pick between a vertical-qualified term and a generic one? Pull both and compare volume against difficulty on the same day. In our case the generic terms won by an order of magnitude on volume at lower difficulty, which is not what we expected before running it.

What about keyword cannibalisation inside a cluster? Give every spoke a distinct primary term and keep the head term out of spoke titles and H1s. Moz's guide↗ covers the basics of intent separation, and the practical check is whether two of your pages ever appear for the same query in Search Console.

Sources and further reading

• Google Search Central, creating helpful content↗ : the who, how and why framing that the first-hand data column implements.
• Moz, Beginner's Guide to SEO↗ : intent separation and internal link fundamentals.
• Search Engine Land↗ : independent coverage for dating algorithm changes referenced in cluster planning.
• Google Search Console Help Center↗ : the cannibalisation check described above.

The keyword figures above are Ahrefs API pulls dated 2026-08-27 for the Taiwan and United States markets. The acceptance pass rates and the cut list are from our own 2026 planning work and are not in the sources listed.

Related: the programmatic SEO playbook covers what happens when you skip the second column, AI Overviews optimization covers why the English click data misleads, and the iGaming SEO service runs this planning pass as its first deliverable.

---

Explore our advertising strategy & media buying services →