Topical Authority

Topical authority is the site-level expertise signal that determines whether search engines and AI systems treat your domain as the trusted source on a specific subject. It is not domain authority. Domain authority measures backlink volume across all topics. Topical authority measures how deeply you cover one subject. A site with a domain rating of 15 can outrank a site with a domain rating of 96 for a topic it owns more completely. That is topical authority doing exactly what it was designed to do.

For years, practitioners debated whether topical authority was a real ranking factor or an SEO folk theory built on correlation. That debate ended in May 2024 when Google's internal API documentation leaked, revealing named production attributes that quantify exactly how deeply a domain covers a subject. The signals exist. They are named. They are in the production infrastructure. What remains is understanding how they work, how they compound, and why they now matter more for AI citation than for traditional search ranking.

What Google's API leak confirmed

The 2024 API leak exposed 2,596 modules containing 14,014 attributes across 2,500+ pages of internal documentation. Among them: three signals that directly measure topical coherence at the site level.

siteFocusScore quantifies how dedicated a site is to a single topic. Specialist versus generalist. A site covering one subject deeply scores higher than a site covering twenty subjects thinly.

siteRadius measures how much an individual page deviates from the site's central theme. Every off-topic page increases the radius and weakens the topical signal.

site2vecEmbeddingEncoded is a compressed vector embedding of the entire site's theme, with pageEmbeddings measuring each individual page against that site-level vector.

These are not the only signals. Fahlout's research synthesis identified at least seven overlapping systems, including NsrChunks (independent quality evaluation per topical section), ClusterUplift (collective quality boosts applied to groups of similar sites), and QualityAuthorityTopicEmbeddings (multi-dimensional vector positioning relative to all other sites). This is not one toggleable score. It is the combined output of topic embeddings, site-level quality chunks, cluster dynamics, and pairwise quality comparisons.

iPullRank's analysis and Hobo Web's breakdown confirmed the practical implication: publishing 50 thin articles on scattered topics does not move the topic embedding vector. Twenty deeply researched, entity-rich, properly interlinked articles on one subject will. Depth and entity coverage shift the vector. Volume alone does not.

Why topical authority matters more for AI citation than traditional search

Traditional search engines rank pages. AI systems rank sources. That distinction changes what topical authority does.

When Google's algorithm evaluates a page, it considers backlinks, keyword relevance, and technical factors at the page level. When an LLM generates an answer, it draws from patterns learned across entire sites, evaluating whether a source consistently demonstrates expertise on a topic. The page still matters. The site behind it determines whether the AI trusts it enough to cite.

The data backs this up. Research analyzing 36 million AI Overviews found a 0.77 Spearman correlation between fan-out query coverage and AI citation likelihood. Fan-out coverage means the number of related sub-queries a site surfaces for within a topic. That is a direct measurement of topical depth, and a 0.77 correlation is one of the strongest ranking signals ever publicly documented.

Text relevance, the strongest individual ranking factor at 0.47 correlation across 16,298 keywords, still operates at the page level. But fan-out coverage operates at the site level. Together they form the two-layer evaluation: the page answers the question, and the site proves the page has the right to be believed.

Analysis of 57,000+ URLs cited in AI Overviews by Surfer SEO found that cited pages have a 29% higher fact coverage ratio than non-cited pages. Fact coverage is the share of facts and subtopics related to the query that a page addresses. Sites with topical authority produce pages with higher fact coverage because the depth already exists in the surrounding content. The page can reference internal cluster pieces, ground claims in adjacent articles, and present the full scope of an answer because the site owns the full scope of the topic.

Ahrefs documented the compounding effect through Healthline, a site whose authority on health topics means a single page about magnesium glycinate ranks for 2,500 keywords on Google, appears in 473 AI Overview queries, 279 ChatGPT prompts, 200 Perplexity prompts, and 86 Copilot prompts. That is topical authority converting into multi-engine visibility at a scale no single-page optimization can replicate.

How topical authority accelerates visibility

Graphite conducted a study across 12 websites and 332 URLs, segmenting content into three buckets by topical authority score. The findings were unambiguous.

Content published on sites with high topical authority (scores above 80) gained its first impression and first click significantly faster than content on lower-authority sites. High-authority content reached impression milestones at 2,500, 5,000, and 7,500 impressions faster across the board. The velocity difference was measurable and consistent: high topical authority content achieved visibility 57% faster than low-authority content.

The compounding mechanism is the practical payoff. Each piece of content published within a topic cluster benefits from the authority already built by surrounding pages. A new article on a site with 30 existing articles in the same topic enters the ranking pool with the site's topical credibility already attached to it. A new article on a site with one article in that topic enters with nothing.

This is why algorithm updates hit broad, shallow content strategies the hardest. Google publicly confirmed in May 2023 that its topic authority system evaluates notability, influence through original reporting, and source reputation when surfacing expert content. The system is designed to reward sites that demonstrate genuine expertise over sites that publish high-volume, low-depth coverage across many topics. During core updates, that reward gets recalibrated. Sites with real topical authority tend to hold their positions. Sites faking it with thin content and broad coverage tend to lose theirs.

Topical authority versus domain authority

Domain authority is a backlink-based metric. It measures how many external sites link to your domain. A site can have high domain authority from backlinks spread across dozens of unrelated topics without being an authority on any single one of them.

Topical authority is narrower, deeper, and increasingly more valuable. Ahrefs documented this through Bicycle Motor Works, a specialist e-bike retailer with a domain rating of 15 that outranks Amazon (domain rating of 96) for competitive e-bike keywords. It also earns regular AI Overview appearances because it owns the topic more completely than a generalist with a diluted focus.

Dimension	Domain Authority	Topical Authority
What it measures	Backlink strength across all topics	Content depth on a specific subject
Signal type	Link-based, domain-wide	Content-based, topic-specific
AI citation impact	Weak correlation	0.77 Spearman correlation with AI citation likelihood
Can a small site win?	Rarely, because backlink volume scales with resources	Consistently, because depth scales with focus
Core update resilience	Volatile when links come from unrelated sources	Stable when coverage is genuine and deep

The practical implication: if you are choosing between acquiring 100 backlinks from unrelated domains and publishing 10 deeply sourced, interlinked articles on one topic, the articles will produce more durable search ranking and significantly more AI citation eligibility. The backlinks move a number. The articles move the embedding vector.

Building topical authority that compounds

Topical authority is built through five operational layers, each reinforcing the others.

1. Topic cluster architecture. Define the subject you intend to own. Map every sub-query, related entity, and angle a searcher or AI engine might explore within that topic. Build content for each one. Interlink them so the site's internal structure mirrors the topic's conceptual structure. This is what shifts the site2vecEmbedding and keeps siteRadius tight.
2. Entity density. Each piece must include named entities: specific companies, people, products, data points, research papers, and industry terms relevant to the topic. Entity chains are the structural mechanism. Every named entity that appears consistently across your topic cluster strengthens the association between your site and that entity in both search indexes and LLM training data.
3. Factual depth over content volume. Twenty articles with original research, specific data, and primary-source citations will outperform 200 thin articles restating the same generic advice. The API leak confirmed this: content depth and entity coverage shift the embedding vector, volume alone does not. Each article must contain claims that are verifiable, specific, and grounded in evidence.
4. Source credibility per piece. Topical authority opens the door. The credibility of each individual page determines whether it gets cited. Named authors, verifiable claims, structural extractability, and answer-first formatting are the per-page requirements. Authority without per-piece credibility is a reputation that never converts to citations.
5. Consistent publishing cadence. Content freshness is a signal AI engines use during retrieval. A site that published deeply on a topic two years ago and stopped signals abandonment. A site that continues to publish, update, and expand its coverage signals ongoing expertise. The cadence proves the authority is current, not historical.

The compounding mechanism is the same one that makes citation architecture work. Each new piece published within the topic cluster benefits from the authority of every piece that came before it. Each citation earned by any piece in the cluster strengthens the site's credibility for every other piece. The flywheel starts slow, accelerates as the coverage deepens, and becomes the competitive moat that generalist competitors cannot replicate without the same focused investment.

Topical authority in Machine Relations

In the Machine Relations framework, topical authority is the foundation layer. Without it, nothing else works.

Entity chains only compound when the site behind them has proven topical depth. Share of citation only grows when AI engines trust the source enough to cite it. AI visibility only becomes durable when the brand is recognized as the expert source on a defined subject, not just a page that answered one question well.

I have watched this play out across dozens of brand placements. The brands that earn consistent AI citations are the ones that own topics, not keywords. They build extractable content around a defined subject, link it internally through entity-optimized structures, and reinforce it through earned media on the same subject from independent publications. The AI engine evaluates the entire surface. If the surface is deep and consistent, the citation follows. If the surface is broad and shallow, it does not.

The question is not whether your brand has content. It is whether your brand owns a topic so thoroughly that an AI engine has no better option than to cite you when that topic comes up. That is the standard topical authority sets. Everything else in Machine Relations, from source authority to AI extractability, is either a component of topical authority or a mechanism for converting it into measurable outcomes.

Frequently asked questions

What is topical authority?

Topical authority is the degree to which search engines and AI systems recognize a website as the expert source on a specific subject. It is built through consistent, deep coverage of a topic across multiple interlinked content pieces, not through backlink volume or publishing breadth. Google's 2024 API leak confirmed it as a multi-signal system including siteFocusScore, siteRadius, and site-level topic embeddings.

How is topical authority different from domain authority?

Domain authority is a backlink-based metric measuring link strength across all topics. Topical authority is a content-depth signal measuring expertise in a specific subject area. A specialist site with a domain rating of 15 can outrank a generalist with a domain rating of 96 on topics it covers more deeply. For AI citation, the distinction is even sharper: fan-out topic coverage has a 0.77 Spearman correlation with AI citation likelihood, while backlink-based metrics show weak correlation.

Does topical authority affect AI search citations?

Yes. Research across 36 million AI Overviews found that sites with broader sub-query coverage on a topic are cited at significantly higher rates. Surfer SEO's study of 57,000+ cited URLs found that pages cited in AI Overviews have a 29% higher fact coverage ratio. Topical authority produces the depth that AI retrieval systems look for when selecting sources to cite.

How long does it take to build topical authority?

Graphite's research found that high topical authority content achieves visibility 57% faster than low-authority content. But building the authority itself requires consistent, deep publishing on a single subject over months. The compounding effect accelerates once the site crosses a coverage threshold where each new piece benefits from the authority of every piece already published.

Can a small website build topical authority?

Yes, and small specialist sites often build it faster than large generalists. Focus is the advantage. A site dedicated entirely to one subject will have a tighter siteRadius, higher siteFocusScore, and a more coherent topic embedding than a large site publishing across dozens of unrelated categories. The evidence shows specialist sites consistently outranking larger competitors on their owned topics in both traditional search and AI citations.