Domain Authority Explains Less Than 4% of AI Citation Variance. Topical Authority Explains 10x More.
For roughly a decade, domain authority has been the metric SEO professionals, such as Radiant Elephant, use to predict competitive outcomes. Higher DA means better rankings. Better rankings mean more traffic. The logic was clean, and the metric was measurable.
Then AI search showed up and mostly ignored it.
Wellows' research found that topical authority (measured as the breadth of keywords a domain ranks for within a topic) correlates with AI citation at r=0.41, making it the single strongest individual predictor of whether AI systems cite your content. Domain Authority, the metric we've collectively obsessed over? It explains less than 4% of citation variance (r²=0.032).
Think about that for a second. A decade of chasing DA, and it accounts for less than four percent of the thing we're now trying to optimize for.
Why topical authority predicts AI citations and domain authority doesn't
The reason maps directly to how retrieval-augmented generation works under the hood. When someone asks ChatGPT or Perplexity a question, the system doesn't just run one search. It breaks the query into multiple sub-queries (fan-out) and searches for the best-matching content across each sub-question.
Kevin Indig's research quantified this: 89.6% of ChatGPT queries generate two or more follow-up searches, and 32.9% of cited pages appeared only in fan-out query SERPs. A third of all cited content would never have been found through traditional single-keyword targeting. The AI found it because it was searching for answers to sub-questions the user never explicitly asked.
Sites with comprehensive topic coverage give the AI more citation opportunities across those fan-out sub-queries. A domain with one great page about a topic has one shot at getting cited. A domain with a pillar page plus fifteen supporting articles has sixteen shots, each matching a different sub-query the AI generates.
Domain authority doesn't capture this. A site with DA 80 and one thin page on a topic loses to a site with DA 30 and deep, structured coverage of that same topic. Growth Memo's March 2026 analysis confirmed the compounding effect: the top 10 domains in a given topic cluster take 46% of all ChatGPT citations, and the top 30 take 67%. Once you establish deep topical coverage, the advantage compounds. Late entrants face a steep climb.
Fan-out queries are why topic clusters win
Surfer SEO's study of 10,000 keywords gave us the most direct evidence for why hub-and-spoke content architecture works for AI visibility. Pages ranking for fan-out sub-queries are 161% more likely to be cited in AI Overviews. Fifty-one percent of all AI Overview citations go to pages ranking for both the main query and at least one fan-out query. Under 20% of citations go to pages ranking only for the main query.
Marie Haynes was among the first to document Google's query fan-out mechanism, noting in March 2025 that queries have "ultimately turned into conversations" under this architecture. BrightEdge's data reinforces the structural point: 82.5% of AI citations go to deep, nested topic pages. Only 0.5% cite homepages.
The pattern is consistent. AI systems are not looking for the most authoritative domain in a general sense. They're looking for the best answer to each specific sub-question. And the best way to have the best answer to many sub-questions is to have built content that actually covers many sub-questions.
That's what topical authority is. Comprehensive coverage of a subject, organized so that each piece stands alone as an answer to a specific question while connecting to the broader topic through internal linking.
How to build topic clusters that compound AI citations
Start with a pillar page. This is a comprehensive piece (2,500-5,000 words) covering your core topic. Not a surface-level overview. A genuine deep dive that addresses multiple user intents, includes specific data, and references the sub-topics your spoke pages will cover in detail.
Build 15-20 spoke pages around that pillar, each addressing a specific subtopic, question, or angle. Each spoke should target a different query the AI might generate during fan-out. "What is [topic]?" is one spoke. "How does [topic] compare to [alternative]?" is another. "[Topic] for [specific industry]" is a third.
Use bidirectional internal linking. Every spoke links to the pillar, and the pillar links to every spoke. This creates a content graph that AI systems can traverse, building confidence that your domain covers the full topic, not just one angle of it.
Target the fan-out queries specifically. Don't just optimize for your primary keyword. Figure out what the second, third, and fourth questions a user might ask after the first one, and build content for those. Tools like iPullRank's Qforia (Gemini-powered) and Wellows' Query Fan-Out Generator can simulate the sub-queries AI systems generate, giving you a direct map of what to create.
And prioritize depth over breadth across topics. A domain that covers three topics with deep cluster architecture will outperform one that covers thirty topics with one article each. AI rewards concentration. Spread yourself thin, and you're a generalist. Go deep, and you're an authority.
The brands winning AI citations aren't the ones with the highest DA. They're the ones who own their topic completely. I covered topical authority alongside 14 other evidence-backed GEO tactics in a full research review synthesizing 12 studies and 17 million citations. Click here to read it.
