Domain Authority vs Page Optimization
By Andrew Coffey · 2026-03-24
Everyone in this space knows domain authority matters for AI citation. Ahrefs showed it. SE Ranking showed it. Every SEO practitioner has seen it firsthand. But how much does it matter relative to page optimization? If you had to choose between improving your backlink profile and fixing your on-page structure, which moves the needle more? And does the answer change depending on how authoritative your site already is? Nobody had measured both dimensions from the same dataset — page-level HTML signals and domain-level authority metrics, for the same cited and control pages, with the same statistical methods. So we did.
The Domain Signals
We analyzed 1,000 cited domains (selected by citation frequency — the most-cited first) and 1,000 control domains (randomly sampled from Brave Search results) using DataForSEO's Backlinks and Domain Analytics APIs. The effects are dramatically larger than anything at the page level. Domain Rank (0-1000) at d=1.075 (cited 431, control 277), Referring Subnets d=0.513 (cited 14,700, control 1,800), Referring IPs d=0.404 (cited 33,100, control 3,300), Crawled Pages d=0.334 (cited 345,500, control 53,300), Referring Main Domains d=0.242 (cited 83,100, control 6,600), Referring Domains d=0.235 (cited 96,500, control 7,400), Total Backlinks d=0.098 (cited 48.7M, control 467K). The Domain Rank score (d=1.075) is the single strongest signal in our entire calibration system. Domain authority is roughly 7x more predictive than the best page-level signal. This confirms what Ahrefs found with a different methodology: their brand web mentions showed ρ=0.664 Spearman correlation with AI Overview visibility. SE Ranking found sites with over 32,000 referring domains are 3.5x more likely to be cited by ChatGPT. Different measurements, same conclusion.
Backlink Diversity: The Overlooked Factor
The referring subnets finding (d=0.513) deserves its own section because we haven't seen it published elsewhere. A "referring subnet" is a unique Class C network block (the first three parts of an IP address). If a domain has 10,000 referring domains but they're all hosted on the same server farm, that's a small number of subnets — likely a private blog network or link scheme. If those 10,000 referring domains are spread across 14,700 different subnets, that's genuine independent endorsement from across the internet. Referring subnets is more than twice as predictive as referring domain count (d=0.513 vs d=0.235). Total backlinks (d=0.098) barely registers. The practical implication: ten links from ten different hosting networks are worth more than a hundred links from the same infrastructure. Quality and diversity of endorsement matters more than volume.
Search Visibility: Long-Tail Breadth Beats Top Rankings
The Domain Analytics data revealed something counterintuitive about search visibility: Organic Keywords pos 91-100 d=0.281 (cited 24,700, control 2,300), Organic Keywords pos 81-90 d=0.267 (cited 41,700, control 3,900), Organic Keywords pos 1 d=0.139 (cited 28,000, control 1,200). Keywords ranking in positions 91-100 (d=0.281) are more predictive of citation than keywords ranking in position 1 (d=0.139). This isn't saying that ranking #95 helps you get cited. It's measuring domain breadth. A domain that ranks for 24,700 keywords even at the bottom of search results has massive topical coverage. That breadth is a proxy for comprehensive authority. Ahrefs found that 80% of LLM citations don't even rank in Google's top 100 for the original query. What matters isn't any individual ranking, it's whether the domain has demonstrated comprehensive topical coverage across thousands of queries.
The Combined Analysis: 77% Domain, 23% Page
We joined the page-level and domain-level data. Each page got its 10 page-level composite scores plus its domain's 4 domain-level composite scores. We ran a logistic regression predicting citation status from all 14 features. Feature importance: Link Quality 24.3% (Domain), Search Visibility 23.1% (Domain), Domain Authority 19.4% (Domain), Domain Scale 10.4% (Domain), Domain Expertise 5.4% (Page), Crawl & Index Signals 3.4% (Page), all other page metrics 1.0-2.6% each. Domain factors: 77.2%. Page factors: 22.9%. The variance decomposition showed that domain and page factors measure genuinely independent dimensions. Domain factors alone improve prediction by R²=0.097. Page factors alone improve it by R²=0.001. Combined: R²=0.100. An important caveat: McFadden pseudo R² is structurally lower than OLS R² and cannot be interpreted as "percentage of variance explained." Our 0.10 represents a meaningful improvement over random guessing, but the majority of what determines citation is something we didn't measure — likely whether the content matches the specific question someone asked.
What This Means for Different Kinds of Sites
If you're an established, high-authority domain: Page optimization is worth your time. The quartile analysis shows that among top-quartile domains, page-level signals produce d=0.24-0.37 effects — meaningful and measurable. You're competing against other high-authority domains, and page quality is the tiebreaker. If you're a growing mid-authority domain: Focus on building authority. Publish consistently. Earn mentions and links from diverse, independent sources. SE Ranking found that domains with profiles on Trustpilot, G2, Capterra, and Yelp have 3x higher citation chances. Page optimization is secondary right now, but do the basics so you're ready when your authority catches up. If you're a new or small domain: Be honest with yourself about the timeline. Domain authority doesn't come from a checklist — it comes from years of consistent work. The best thing you can do is publish comprehensive content on your topics, build your topical footprint, and let authority accumulate. AI citation will follow as your domain grows.