Last updated: 22 August 2026. Written by Adrian Tan, Singapore Digital Marketing (SDM).
“Topical authority” is one of those phrases that arrives in a proposal already sounding expensive. Somebody produces a diagram of a pillar page ringed by twenty spokes, the recommendation is a thirty-post content programme, and nobody in the room quite feels able to ask whether Google has ever confirmed that any of this is a thing.
It is worth asking, because the honest answer is awkward. Asked about topical authority on Reddit in November 2023, Google’s John Mueller replied “This” to a comment describing it as relevancy with a fashionable new name, and followed up with “Don’t worry about it.” There is no topical authority score at Google, no label applied to a domain, and no dial anyone can turn.
And yet the pattern the phrase describes keeps working. This guide separates the two: what Google actually rewards, why clusters deliver it, and — the part global advice consistently gets wrong for this market — how big a cluster should be when your entire addressable search demand in Singapore might be thirty topics deep. It is part of our content marketing guide for Singapore businesses.
What Google actually said, and what it means
Mueller’s “don’t worry about it” is not permission to publish disconnected posts. It is a statement that the cluster is not the ranking signal. Building a hub-and-spoke diagram earns nothing on its own. Google does not scan your site, detect a cluster shape and award relevance points.
What Google does say, in its own documentation, is narrower and more useful. Its SEO starter guide notes that “the vast majority of the new pages Google finds every day are through links”, recommends descriptive anchor text so people and its systems understand what is being linked to, and suggests grouping similar content in directories to help Google understand how often different content types change. Its people-first content guidance asks whether a site “has a primary purpose or focus” and whether content offers “substantial, complete, or comprehensive description of the topic”.
Put those together and you get the real claim: a cluster is a delivery mechanism for things Google does reward — coverage, discoverability, clear relevance and site-level focus. It is not a signal in itself. That distinction matters, because it tells you which parts of the work to do properly and which parts are theatre.
The three mechanisms that actually do the work
1. Coverage changes what you can compete for. A single page about “content marketing” competes for one query against sites that have twenty pages on the subject. Not because those sites have an authority score, but because each of their twenty pages independently matches queries yours does not cover, and because a query with a specific answer is easier to win than a broad one. Coverage is not a signal; it is a larger surface area.
2. Internal links do three jobs at once. They get new pages discovered, which Google says is how most new pages are found. Their anchor text tells Google what the destination page is about — often more reliably than the destination’s own copy. And they route attention: a reader who arrives on a spoke and leaves via a link to your service page is worth more than one who leaves via the back button.
3. Site-level focus is real, even if “topical authority” is not. Google’s guidance explicitly asks whether a site has a primary purpose. A domain whose 40 pages all address one subject reads differently from one whose 40 pages address eleven. This is the closest thing to what people mean by topical authority, and it is better described as coherence.
None of the three requires a diagram. All three require decisions about what you write and how you link it.
The anatomy of a cluster
The structure is simple enough to hold in your head. Three layers, three link directions.
Two rules keep this from going wrong. The pillar must not target the same commercial term as the service page — if your service page owns “content marketing agency Singapore”, the pillar targets the informational version of the subject, or the two compete and Google picks one, usually the wrong one. And the pillar’s index must be maintained. A “where to go next” list that stopped being updated in March is the most common reason clusters underperform: half the spokes have no inbound link and were never discovered properly.
Sizing a cluster for Singapore
Here is where imported advice fails. American content strategy assumes a market where a mid-sized subject supports fifty posts with measurable demand. Singapore has 5.78 million internet users at 98.4% penetration, according to DataReportal’s Digital 2026 report for Singapore. Volume scales with population, so the same subject that supports fifty posts in the US might support twelve here with any measurable search volume at all.
That does not mean write twelve. It means the balance shifts: in a small market, a cluster is built for coverage and conversion rather than for traffic volume, and a meaningful share of its posts will be ones a keyword tool cannot see. We routinely publish posts in Singapore clusters with no recorded volume because sales hears the question weekly — and they convert better than the high-volume posts do.
Building your first cluster in ninety days
A realistic sequence for an SME publishing two posts a month, with one person on it part-time.
| Phase | What you do | Output |
|---|---|---|
| Days 1-10 | Pick the subject. It must map to something you sell, and you must have operating experience in it. Harvest every question sales and support actually receive. | A raw list of 30-50 questions |
| Days 10-20 | Group the questions into a pillar, two cornerstones and eight to twelve spokes. Score and order them. Kill anything a four-sentence AI summary would fully answer. | An ordered cluster map |
| Days 20-35 | Write the pillar. It is the broad guide, it defines the subject’s scope, and it carries the index of everything below it. Longest piece in the cluster. | Pillar live |
| Days 35-60 | Write both cornerstones. These are the two sub-topics with the most commercial weight — usually cost and how-to-choose. | Two cornerstones live |
| Days 60-90 | Write spokes in order. Each one: link up, link to pillar, two sibling links, and add it to the pillar index the same day. | Four to six spokes live |
| Day 90 | Audit. Every page has at least two inbound internal links. Nothing is orphaned. Check the query report for overlap. | Cluster ready to extend |
The order matters more than the pace. Publishing eight spokes before the pillar exists means eight pages with nowhere to link and no page consolidating the subject — and it is what most stalled content programmes did.
The internal linking rules that do the work
- Descriptive anchors, always. Google recommends anchor text that tells people and its systems what is being linked to. “Our guide to content marketing ROI” works; “read more” does not.
- Links in body copy, not just in a related-posts block. A link inside a sentence where the reader is already thinking about the topic is used; a block of six links under the footer is not.
- Inbound on day one. This is the rule most often broken. When a post goes live, add it to the pillar index and add one contextual link from an existing sibling that day. Outbound links do not stop a page being an orphan; inbound ones do.
- Two inbound links minimum, forever. Audit for pages with fewer and fix them. A page with one inbound link is one edit away from being orphaned.
- Vary the anchor. The same exact-match phrase on every link to a page reads as mechanical and tells Google less than three natural variants would.
- Link across clusters where the reader would. A content post that references measurement should link to the measurement cluster. Rigid silo purity is a made-up rule that produces worse pages.
Cannibalisation: the failure mode nobody warns you about
Clusters have a specific way of going wrong, and it takes about eight months to show up. You publish thoroughly around a subject; two or three of your posts end up sufficiently similar that Google cannot tell which should rank for a query; all of them sit at position eight instead of one of them sitting at position three.
How to spot it: in Search Console, filter to a query and look at the pages. If two or more of your URLs appear for the same query with similar impressions and neither ranks well, that is cannibalisation.
How to fix it, in order of preference:
- Merge. Fold the weaker page into the stronger one, 301 redirect it, and update the internal links pointing at it. This is right roughly two-thirds of the time.
- Differentiate. If both topics are genuinely distinct, rewrite each so the distinction is obvious in the title, the H1 and the opening paragraph. “SEO cost” and “is SEO worth it” are different questions; write them as different questions.
- Demote. Keep both, but make one clearly subordinate: remove its competing links, point its internal links at the other, and let it serve the long tail.
Prevention is cheaper. Before writing anything, check whether an existing page already ranks between positions four and fifteen for the target query. If it does, extending that page almost always beats publishing a new one — a point covered in more detail in our guide to duplicate and overlapping content.
How to tell whether it is working
Cluster performance is measured at the cluster level, not the post level. Four things to watch, roughly in the order they move:
- Total unique queries the cluster receives impressions for. This is the earliest indicator and it moves within weeks. A cluster gaining query breadth is working even if positions look flat.
- Share of your target query set where you appear in the top twenty. Track the cluster’s thirty target queries as a set. Going from six to eighteen is progress; the average position across them is noise.
- Internal traffic flow to the service page. Sessions arriving at the service page from cluster pages. This is the number that connects the content to revenue.
- Assisted conversions. Enquiries where a cluster page appeared anywhere in the path. Most content assists rather than closes, so last-click reporting will tell you the blog does nothing. Our guides to attribution models and content marketing ROI cover how to see it.
Give a cluster six months before judging it. Give an individual post six weeks before judging whether it is relevant at all.
When a cluster is the wrong move
Three situations where we advise against it.
When the site cannot yet be crawled or trusted. Twenty posts on a site with unresolved technical problems is twenty pages that will not perform. Fix the technical basics first; it is a two-week job, not a two-quarter one.
When you need enquiries within eight weeks. Clusters compound over quarters. If the business needs pipeline this month, that is a paid media problem — see SEO versus Google Ads — and content is what you build alongside it, not instead of it.
When you cannot sustain the cadence. Six well-linked posts and a maintained index beat a half-built twenty-post plan. Decide the size against the hours you actually have, using a publishing cadence you can hold.
Frequently asked questions
Is topical authority a real Google ranking factor?
No. There is no topical authority score and no label Google applies to a domain. Asked about it on Reddit in November 2023, Google’s John Mueller endorsed a comment describing the term as relevancy rebranded, and said “Don’t worry about it.” What is real is what the term gestures at: coverage that lets you compete for more queries, internal links that get pages discovered and understood, and site-level focus, which Google’s own guidance touches on when it asks whether a site has a primary purpose. Build for those three and you get the outcome without needing the label.
How many posts does a content cluster need?
In Singapore, usually eight to fifteen for a single subject — a pillar, two cornerstones and five to twelve spokes. The number is set by how many genuinely distinct questions exist, not by a target. With 5.78 million internet users, many valid Singapore topics have little or no measurable search volume, so expect roughly a third of a good cluster to be posts a keyword tool cannot see. Write them if sales hears the question; they often convert better than the high-volume ones.
Should the pillar page or the service page target the commercial keyword?
The service page, always. If your service page targets “content marketing agency Singapore”, the pillar must target the informational version of the subject instead. Pointing both at the same commercial term makes them compete, and Google will pick one — frequently the blog post, which converts worse. The pillar’s job is to own the broad informational query, hold the index of everything below it, and pass readers up to the service page with a keyword-rich anchor.
How long before a cluster produces results?
Query breadth usually moves within four to eight weeks of the pillar going live. Rankings on target queries typically take three to six months, and the traffic-to-enquiry effect usually shows around months six to nine, because most content assists conversions rather than closing them. Judge the cluster at six months, not the individual posts. If nothing has moved on query breadth after eight weeks, the problem is relevance or discoverability, not patience — check for orphaned pages first.
What is keyword cannibalisation and how do I know if I have it?
It is when two or more of your pages compete for the same query, so none of them ranks as well as one page would. Find it in Search Console: filter to a query and look at which of your pages receive impressions for it. Two URLs with similar impressions and neither ranking well is the signature. Fix it by merging the weaker page into the stronger one with a 301 redirect and updating internal links, which is the right answer about two-thirds of the time. Otherwise, rewrite both so the distinction is obvious in the title, H1 and opening paragraph.
Do I have to finish one cluster before starting another?
For most Singapore SMEs, yes. A finished cluster of ten posts with a maintained index and no orphans outperforms two half-built clusters of five, because coverage of a subject is what does the work and half-coverage does not deliver it. The exception is a business with genuinely separate audiences whose sales cycles cannot wait — then run two, but staff them separately, because splitting one person across two clusters usually finishes neither.
Where to go next
To choose the subject and order the posts, content keyword research for Singapore covers the scoring model, and how often to publish sizes the plan against your actual hours. For the individual pieces, how to write a blog post that ranks is the method and E-E-A-T for Singapore businesses is how you make each one credible. Once posts are live, repurposing them extends their reach, and earning backlinks is what turns a well-built cluster into one that outranks bigger sites.
If you would rather have the cluster mapped, written and linked for you, that is what our content marketing services in Singapore do — and you can see the results for other Singapore businesses in our case studies.
Sources: Google Search Central, SEO Starter Guide and Creating helpful, reliable, people-first content; Search Engine Journal, Google on topical authority: don’t worry about it (John Mueller, November 2023); DataReportal Digital 2026: Singapore (5.78m internet users, 98.4% penetration, October 2025 data); Google Search Status Dashboard. On grants: ongoing marketing retainers and ad spend are generally not claimable; only pre-approved solutions under the Productivity Solutions Grant are, SDM is a pre-approved PSG vendor, and clients apply for and manage the grant themselves. Enterprise Singapore has said the EDGE framework will consolidate EDG, PSG and MRA from the second half of 2026.
