984 Sites, 12 Industries: What Actually Predicts AI Citations?

When we built the AgentReady scoring framework, we made a bet: that the factors we measure — schema markup, bot access, content quality, AI protocols, authority signals, crawl efficiency — actually correlate with whether AI systems cite a website. That's the whole thesis. And we needed to test it.

So we ran a proper study. 984 websites across 12 industries. We scored each one, then used a panel of 87 AI-generated queries to check which sites actually appeared in AI responses (ChatGPT, Perplexity, and Claude). We calculated Spearman rank correlation coefficients between readiness scores and citation rates for each industry.

The results were more nuanced than we expected — and we're going to share them with complete transparency, including the parts that complicated our thesis.

Across all 984 sites and all 12 industries combined, the Spearman correlation between AI readiness score and citation rate is r=0.025. That's statistically negligible.

If you stopped there, you'd conclude our scoring framework predicts nothing. But stopping there would be wrong — because the aggregate masks dramatic industry-level variation. When we broke out the data by sector, the picture changed completely.

The key insight from our v3 study is this: current AI citations are driven primarily by brand recognition (how well-known a site is from training data), not by technical readiness. The more established a brand, the more AI systems cite it — regardless of schema markup, llms.txt, or crawl optimization. This is the Brand Fame Effect, and it explains why Wikipedia, WebMD, and Investopedia get cited constantly despite not being technically 'AI-ready' by our criteria.

But brand recognition isn't the whole story either.

When we isolated YMYL (Your Money or Your Life) industries — sectors where AI systems apply extra source quality scrutiny — the correlations jumped dramatically.

Healthcare: ρ=0.72 — By far the strongest. Medical sites with strong E-E-A-T signals (author credentials, clinical citations, schema markup) are cited significantly more often than technically weaker peers. AI systems are most cautious about medical misinformation, so they reward trust signals.

Government: ρ=0.40 — Second highest. Government sites with clear entity signals, structured service data, and accessible content are cited more reliably. Regulatory authority matters.

Education: ρ=0.35 — Strong in the .edu domain, where institutional credibility intersects with technical readiness.

Insurance: ρ=0.33 — Similar pattern to healthcare, with policy structure and trust signals driving citation rates.

Finance: ρ=0.20 — Meaningful but more modest, reflecting the complexity of financial information and AI systems' caution about specific recommendations.

For all other industries (tech, e-commerce, media, travel, etc.), correlations were below r=0.15 — weak enough that brand recognition likely explains the residual.

Even with a weak overall correlation, we found one consistent pattern across all industries: sites scoring 95+ were cited 59% of the time, compared to 41% for sites scoring below 50. That's a meaningful gap.

This likely reflects a floor effect: once a site is technically well-optimized, brand recognition amplifies citation rates. The highest-scoring sites tend to be well-known brands that have also invested in technical optimization — a compound advantage.

For less well-known brands, the implication is clear: technical readiness alone won't overcome brand anonymity in AI citations today. But as AI systems shift from training-data citations to real-time crawl-based citation (already happening with Perplexity and Google AI Overviews), technical readiness will become the differentiator.

The strong correlations in healthcare, government, and education aren't accidental. AI systems apply a quality filter called E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) when generating responses in high-stakes domains. That filter is more legible for AI systems when sites have structured signals — author credentials, institutional affiliations, citations, schema markup.

In other words: in YMYL industries, technical readiness acts as a proxy for credibility. Schema markup isn't just a crawlability signal — it's a trust signal. E-E-A-T signals aren't just SEO factors — they're the machine-readable version of authority that AI citation systems rely on.

For brands in healthcare, government, education, insurance, and finance: technical AI readiness is already delivering measurable citation advantages. The correlation data makes this actionable, not hypothetical.

The study doesn't invalidate technical AI readiness — it contextualizes it. Here's what we conclude:

If you're in a YMYL industry: AI readiness improvements have a direct, measurable impact on citation rates today. Healthcare, government, education, insurance, and finance organizations should treat AI readiness as a top-priority initiative.

If you're in a non-YMYL industry: Technical readiness is about positioning for the future, not capturing citations today. As AI systems shift to real-time crawl-based citation, brand recognition will matter less and technical signals will matter more. Sites investing now will have a structural advantage when that shift accelerates.

For everyone: The 95+ score premium suggests that once brand recognition and technical optimization combine, citation rates jump meaningfully. Smaller brands that achieve technical excellence may punch above their weight — especially in industries where dominant brands don't maintain clean technical signals.

You can explore the full dataset and download the raw correlation data on our /research page.