The Complete Ranking: Every Factor That Predicts AI Citation

By Andrew Coffey · 2026-03-24

Posts 1 and 2 looked at page-level and domain-level factors separately. This post puts them together: one table ranking all 14 metrics by measured effect, plus two analyses that test whether page optimization matters independently of domain authority.

The Unified Ranking

We joined every page to its domain's data and ran Cohen's d on all 14 metrics — 10 page composites and 4 domain composites — between cited and control pages. Rank 1: Search Visibility (Domain) d=1.205. Rank 2: Link Quality (Domain) d=1.195. Rank 3: Domain Authority (Domain) d=1.189. Rank 4: Domain Scale (Domain) d=0.796. Rank 5: Crawl & Index Signals (Page) d=0.158, weight 14.8%. Rank 6: RAG Retrieval Suitability (Page) d=0.140, weight 13.0%. Rank 7: Content Relevance (Page) d=0.136, weight 12.6%. Rank 8: Engagement Cues (Page) d=0.135, weight 12.5%. Rank 9: Multimodal Readiness (Page) d=0.123, weight 11.4%. Rank 10: AI Readability (Page) d=0.095, weight 8.8%. Rank 11: Citation Suitability (Page) d=0.077, weight 7.2%. Rank 12: Structured Metadata (Page) d=0.072, weight 6.7%. Rank 13: Page Freshness (Page) d=0.072, weight 6.7%. Rank 14: Domain Expertise (Page) d=0.068, weight 6.3%. The gap between the domain and page tiers is stark. The weakest domain metric (Domain Scale, d=0.796) is 5x stronger than the strongest page metric (Crawl & Index Signals, d=0.158). The top three domain metrics (d=1.189-1.205) are nearly identical to each other, suggesting they're all measuring overlapping aspects of the same thing: how big and well-known is this domain. Within the page tier, the distribution is relatively flat — the #5 metric (d=0.158) is only about 2.3x stronger than the #14 metric (d=0.068). This means page optimization is a "do everything a little" game rather than a "nail one thing" game.

The Quartile Test: Does Page Optimization Work for Everyone?

This is the analysis we haven't seen published anywhere else. Instead of asking "do page signals matter?" it asks "do page signals matter equally across different authority levels?" We split all pages into four groups by their domain's Domain Rank. Then we computed page-level Cohen's d within each group separately. Q1 (lowest, Rank 0-400): Citation Rate 61.7%, Avg Page d ≈ 0.00. Q2 (Rank 400-526): Citation Rate 79.0%, Avg Page d ≈ -0.07. Q3 (Rank 526-611): Citation Rate 75.9%, Avg Page d ≈ -0.17. Q4 (highest, Rank 611-964): Citation Rate 77.6%, Avg Page d ≈ +0.33. The individual metric breakdown by quartile: Content Relevance Q1 +0.01, Q2 -0.09, Q3 -0.22, Q4 +0.36. Crawl & Index Signals Q1 +0.02, Q2 +0.01, Q3 -0.07, Q4 +0.37. Engagement Cues Q1 -0.04, Q2 -0.08, Q3 -0.15, Q4 +0.36. Multimodal Readiness Q1 -0.00, Q2 -0.10, Q3 -0.14, Q4 +0.34. RAG Retrieval Suitability Q1 -0.02, Q2 -0.03, Q3 -0.15, Q4 +0.33. Citation Suitability Q1 +0.08, Q2 -0.07, Q3 -0.29, Q4 +0.33. AI Readability Q1 +0.04, Q2 -0.06, Q3 -0.29, Q4 +0.32. Domain Expertise Q1 -0.01, Q2 -0.09, Q3 -0.25, Q4 +0.28. Page Freshness Q1 -0.08, Q2 -0.09, Q3 -0.08, Q4 +0.25. Structured Metadata Q1 +0.00, Q2 -0.11, Q3 -0.09, Q4 +0.24. Every single page metric is positive in Q4. Every single one is flat or negative in Q1-Q3. This is a clear pattern: page optimization compounds with domain authority. Among the top 25% of domains by authority, well-optimized pages are measurably more likely to be cited than poorly-optimized pages on similarly authoritative domains. Among the bottom 75%, page structure doesn't measurably differentiate cited from non-cited pages. What this means practically: If your domain is in Q4 (DataForSEO Rank above ~611, roughly equivalent to high-authority domains with diverse backlink profiles), every page-level improvement is worth doing. The effects are d=0.24-0.37 — meaningful and consistent across all metrics. If your domain is in Q1-Q3, page-level changes won't produce measurable citation improvements today. That doesn't mean they're worthless — they prepare your pages for the day your authority catches up, and they improve user experience regardless. But be realistic about what they'll accomplish for AI citation in the short term.

The Within-Domain Test: Holding Authority Perfectly Constant

The quartile analysis controls for authority approximately. But 510 domains in our dataset have both cited AND non-cited pages — meaning domain authority is literally identical between the two groups because they're on the same domain. We filtered to 272 qualifying domains (those with at least 2 cited and 2 non-cited pages) and ran Cohen's d on the 10 page metrics. Crawl & Index Signals: within-domain d=+0.103 (overall d=+0.158). RAG Retrieval Suitability: +0.076 (overall +0.140). Engagement Cues: +0.076 (overall +0.135). Multimodal Readiness: +0.072 (overall +0.123). Content Relevance: +0.058 (overall +0.136). Page Freshness: +0.052 (overall +0.072). Structured Metadata: +0.046 (overall +0.072). Citation Suitability: +0.029 (overall +0.077). AI Readability: +0.026 (overall +0.095). Domain Expertise: +0.018 (overall +0.068). Every metric is positive. The effects are small — only Crawl & Index Signals rises above "table stakes" at d=0.103 — but they're consistent. This confirms two things simultaneously: On-page signals have real, independent predictive value — they're not just proxies for domain quality. And the independent effect is much smaller than the domain-mediated effect — the within-domain d values are roughly 40-60% of the overall d values.

The Honest Conclusion

Domain authority is the gatekeeper. Page optimization is the tiebreaker. Both are real. Neither is sufficient alone. And the majority of what determines citation — whether your content matches the question someone asked — isn't captured by either. If you want one actionable sentence: build your domain's authority through consistent, comprehensive content and genuine brand-building, and make sure your pages have clean HTML fundamentals so you're ready when that authority kicks in.