Topical authority, without the snake oil.

“Topical authority” is the SEO industry’s current favourite phrase. It will be replaced inside two years — the names always are — but the underlying mechanic isn’t new and isn’t complicated. Search engines reward sites that demonstrably know more than other sites about a coherent set of topics, and they punish sites that pretend to. This essay is about what demonstrating that, in practice, looks like.

What “topical authority” isn’t

It isn’t writing 200 articles about a single keyword cluster. It isn’t internal-linking discipline alone. It isn’t a Schema.org markup configuration. All of these come up when authority is discussed and all of them are downstream tactics, not the thing itself.

What it actually is

Topical authority, operationally, is the cumulative effect of (a) the breadth and depth of your published content on a topic, (b) the quality of internal information architecture around that content, (c) external citations and links into the content from sources that themselves have authority on the same topic, and (d) the absence of low-quality content under the same domain that dilutes the signal. The first three you build deliberately; the fourth is a hygiene exercise most teams skip.

The pillar-and-cluster trap

The pillar-and-cluster model — one long “pillar” page on a topic, surrounded by a cluster of supporting articles linking to it — is genuinely useful as a structural framework. It is also massively over-applied. We see clients with twenty “pillar” pages, each surrounded by clusters of indistinguishable supporting content, none of which would be missed if it disappeared overnight. The pillar-and-cluster pattern works when each pillar is genuinely the canonical resource on its topic; it fails when it’s a content-production scaffolding repurposed as a publishing strategy.

Coverage breadth versus coverage depth

The most useful framing we’ve found: are you trying to cover a topic comprehensively, or own it definitively? Coverage means publishing across the full range of subtopics so that any informational query in the space lands somewhere on your site. Ownership means building the resource that the rest of the industry refers to when discussing one specific subtopic. Most clients try to do both at once and end up with neither.

The right move depends on your competitive position. If you’re a category challenger, pick three or four subtopics where the incumbents are weak and own them definitively. If you’re an incumbent, your job is coverage breadth: don’t let challengers find any subtopic of the parent topic where they can outrank you.

Where AI-generated content actually fits

It does fit. We have used AI-assisted drafting on roughly a third of the content we’ve published in the last twelve months. The pattern that works: a senior writer briefs the article, an AI tool produces a structured first draft, the writer rewrites the draft significantly, the editor reviews against an originality benchmark before publishing. Time saved versus a from-scratch first draft is around 30%, which compounds over a publishing programme. What does not work: AI tool produces an article, junior reviewer applies cosmetic edits, article goes live. We have seen accounts lose meaningful traffic to a single Google update after publishing 30 to 50 of these in succession.

The content-pruning dividend

The boring intervention with the largest impact-per-effort ratio in our SEO work is auditing and pruning content. A typical takeover client has between 15% and 40% of their indexed pages contributing zero organic traffic and not earning a meaningful link. These pages dilute topical authority signals; they also slow crawl and waste internal-linking budget. We routinely see organic traffic increase by 12% to 25% within three months of a disciplined pruning exercise — not because the remaining content suddenly improves, but because the noise floor drops.

The metric we actually track

Topical authority is hard to measure in isolation. The proxy metric we use across client engagements is share-of-voice on a defined keyword universe, segmented by topic cluster. We define the keyword universe at engagement start (typically 800 to 2,000 informational and commercial-intent queries), pull weekly ranking data for every query, and aggregate to cluster-level share of voice. A successful programme moves cluster-level share-of-voice up over six to nine months without spiking and reverting.

What changed in 2025

The single biggest practical change of the last twelve months is the visible weight Google now puts on first-party data and verifiable expertise. The structured-data work that matters most now is reliable Author, Organization and Article markup with verifiable links to the author’s public profile and credentials. Anonymous content publishes still rank but the floor on what passes a quality review has clearly moved up. If your in-house authors don’t have a public footprint, build them one.

Closing

The mechanic hasn’t changed. Search engines reward demonstrably knowing your subject. The tactics keep getting renamed, but the work is the same: publish things only experts could write, organise them so a reader can navigate the topic from any entry point, get cited by other authorities, and don’t pollute your domain with content you wouldn’t personally show a knowledgeable peer. Everything else is implementation detail.