The cluster was complete. The authority was not.

Meta description: When every AI-written page sounds the same, authority comes from specific, verifiable, firsthand knowledge — not from more coverage.

The SEO teams I meet in this situation are not amateurs. They did what the last decade taught, and they did it well: a pillar page that owns the topic, a cluster of supporting articles, internal links arranged like a well-kept garden. The cluster is complete by every criterion the discipline ever published. And the AI Overviews on their core questions cite a competitor, or a forum, or a PDF from an association. The team concludes the cluster needs refreshing. I have come to read something else: the authority itself has come apart, because what authority rests on has changed.

A business can publish a great deal of content and still become less distinctive. Articles in the same industry often follow the same structure, repeat the same generic advice, and produce the same polished but interchangeable answers. Increasing production does not necessarily increase the amount of information that a potential customer — or an AI-assisted search product — can use to tell one provider from another.

What did the old authority model assume about retrieval?

The old model assumed a crawler that indexed pages and a ranker that rewarded demonstrated coverage. Cover a topic thoroughly, link it coherently, earn references, and the system reads that as authority. That assumption held for a retrieval system that returned pages. The answer layer does something different: it assembles an answer from passages, and it favors the passage that resolves the specific sub-question being asked. Authority in that system is not a property of a page collection. It is a property of something the cluster model never had to be.

Where does that assumption fail when answers are assembled?

My starting question is not how many more pages we can publish. It is what this particular organization knows that a competitor cannot simply reproduce by asking a model for another article. That may be an operational detail, an unusual customer problem, a firsthand observation, a process limitation, or a small but revealing distinction in how a service is delivered. These details can appear minor to a traditional content production team. From my perspective, they are often where genuine differentiation begins.

Over more than twenty years in SEO, I have watched tactics become templates. Once enough businesses imitate a successful format, the format itself stops being a meaningful competitive advantage.

I can describe where the failure shows up. The pillar page answers the general question well, but the answer layer is rarely asked the general question — buyers ask narrow, situational things, and the engine fans the query out into sub-questions. The company’s narrow pages exist, but they were written as spokes: they point back to the pillar, they summarize, they cover. A spoke that covers a sub-question loses, in the cases I watch, to a page that resolves it — even when that page belongs to a source with a fraction of the company’s domain strength.

Why does adding spokes dilute the signal?

The false explanation is coverage: more spokes, more long-tail, more freshness. I have watched clusters double in size while citations kept falling. Each new spoke added words about the topic without adding anything the engine could not get elsewhere, and the mass of interchangeable pages made it harder, not easier, for the answer layer to locate what the company distinctively knows. The team was maintaining a garden. The engine was looking for a source, and the garden had begun to hide it.

What pattern shows authority coming apart rather than “needing freshness”?

Freshness problems look like decay: positions slip gradually, older pages fall first. Authority coming apart looks different in the cases I have watched: the pages hold their classic rankings while citations drain away, and the citations go to sources that are often smaller, narrower, and more specific. The team experiences this as randomness — the engine’s whims. It is not random. The answer layer is selecting for something the cluster was never built to have, and “freshness” is the name the old model gives to a failure it cannot describe.

A hundred similar pages may communicate less about a company than ten carefully constructed pages that each answer a real question from direct experience. I would still maintain the technical foundations and distribution work that made those pages discoverable. But I would not confuse discoverability with a reason to be chosen. The latter requires a clearer account of what only this business can contribute.

An editorial team can compare a generic page with a version that supplies accurate, concrete expertise, inspect the resulting user behavior and search visibility over time, and repeat carefully framed AI queries. Such checks are imperfect. The point is to develop better evidence, not to pretend we have decoded a proprietary system.

[AWAITING MIKLOS — scene: second scene]

What question replaces “do we cover the topic”?

“Do we cover the topic?” is a question about the company and its pages. The teams that ask it produce coverage reports, gap analyses, refresh calendars — all company-facing. There is a question that faces the other direction, toward the buyer and the assembled answer, and I have noticed it reorganizes everything the team does next. It also cannot be answered with a crawl.

The question that replaces “do we cover the topic?” is the one coverage cannot answer: what does this organization know — specifically, verifiably, from direct experience — that a competitor cannot reproduce by asking a model for another article?

PR excerpt (80–120 words, Miklos’s own words): A business can publish a great deal of content and still become less distinctive. Articles in the same industry follow the same structure, repeat the same generic advice, and produce the same polished but interchangeable answers. A hundred similar pages may communicate less about a company than ten carefully constructed pages that each answer a real question from direct experience. Discoverability is not a reason to be chosen — that requires a clear account of what only this business can contribute.

Miklos Roth, AI marketing consultant, Budapest

By Miklos Roth — AI marketing consultant, Budapest. More: his field notes.

© Copyright Teppichreinigung