What Drives AI Citations? Data from 129,000 Domains

SE Ranking analyzed 129,000 domains to find what makes ChatGPT cite a source. Referring domains, content freshness, and page speed matter most, while FAQ schema doesn't.

Key Takeaways

The Study: 129,000 Domains Analyzed

SE Ranking's 2025 study is the largest analysis of ChatGPT citation patterns to date. The researchers examined what characteristics correlate with being cited by AI assistants, the mechanics behind what AEO is in practice.

Why this matters: Most AEO advice is based on assumptions or small-sample observations. This study provides actual data on what works, and what doesn't.

Top Factor: Referring Domains

Backlinks emerged as the strongest predictor of AI citations. Sites with more referring domains are dramatically more likely to be cited.

Referring Domains Avg. Citations Citation Likelihood
350,000+ 8.4 citations Highest
32,000+ ~5-6 citations 3.5x more likely
<2,500 1.6-1.8 citations Baseline

The implication: Link building isn't just for SEO. Sites with strong backlink profiles, built through PR, guest posting, and creating linkable content, also perform better in AI search.

This makes sense: AI models are trained on web data, and backlinks have always been a signal of authority and trustworthiness. The models learned these patterns. It's a theme that came up again in the takeaways from BrightonSEO October 2025, where speakers argued that visibility is fundamentally a byproduct of earned trust.

Content Freshness: The 10-Month Window

AI heavily favors recent content. The data is striking:

  • 95% of ChatGPT citations come from content published in the last 10 months Source: AirOps 2025 Study
  • 76.4% of most-cited pages were updated in the last 30 days Source: Digitaloft Research

The "Last Updated" Timestamp Effect

Pages with a visible "last updated" or "last modified" timestamp receive 1.8x more citations than pages without one. This is a quick win: add a timestamp and keep it current.

What About FAQ Schema? (The Myth Busted)

Many AEO guides recommend FAQ schema as a key tactic. The data tells a different story.

Our own 500-query study reached the same verdict independently: in Every AI Engine Plays by a Different Rulebook, FAQ-marked pages were cited at the same roughly 9% rate by all three engines, so FAQ schema separated nothing, even while schema overall predicted ChatGPT citations strongly.

Pages without FAQ schema received more citations. Why?

The researchers note that FAQ schema often appears on simpler support pages, product FAQs, help center articles, that naturally earn fewer citations than in-depth content. The schema itself isn't hurting citations; it's just correlated with lower-authority page types.

Takeaway

FAQ schema helps AI understand your content structure, but it's not a citation-boosting hack. Focus on the fundamentals: authority, freshness, and content quality. For the full strategy playbook, see our complete AEO guide.

Traffic, Length, and Page Speed

Several other factors showed strong correlations with citation rates:

Factor High Performers Low Performers
Domain Traffic 10M+ visitors = 8.5 citations Low traffic = ~2 citations
Content Length 2,900+ words = 5.1 citations <800 words = 3.2 citations
Page Speed (FCP) <0.4s = 6.7 citations >1.13s = 2.1 citations

All three figures come from SE Ranking's analysis of 129,000 domains and 216,524 pages. They are correlations, not measured causes.

Page speed likely matters because AI crawlers have timeouts. Slow pages may not fully load before the crawler moves on. Faster pages = more content captured = higher citation potential.

The 80% Disconnect: AI Citations vs Google Rankings

Perhaps the most striking finding: AI citations and Google rankings don't overlap as much as you'd expect.

29%of AI citations come from Google Page 1 results (top 10) Source: Digital Bloom 2025
80%of AI-cited sources are NOT in Google's top 3 results Source: Digital Bloom 2025

Seer Interactive found a 0.65 correlation between organic rankings and LLM mentions, positive, but not strong. Ranking #1 on Google doesn't guarantee AI citations. Being cited by AI doesn't require ranking on Google.

A Second Study Says the Battle Is Mostly Off Your Site

SE Ranking's data is not alone anymore. A 2026 analysis by AirOps and Nobori found that roughly 85% of AI visibility does not come from a brand's own website: the engines assemble answers mostly from third party sources such as review platforms, comparison articles and industry roundups.

Read together, the two studies describe one mechanism from two sides. Referring domains predict citations (SE Ranking) because links are how the engines find and trust the third party evidence layer where most citations actually originate (AirOps and Nobori). The practical order of work follows directly: earn presence on the sources the engines already consult for your category first, then make your own pages extractable enough to be quoted alongside them.

This matches what we see in our own client tracking at Novastacks: when we baseline a brand's share of AI answers across a fixed prompt set, the brands with a strong third party footprint start ahead of brands with stronger websites but no external evidence.

What Actually Works: Summary

High Impact

  • Build referring domains (backlinks)
  • Keep content fresh (update within 10 months)
  • Add "last updated" timestamps
  • Improve page speed (FCP <0.4s)
  • Write comprehensive content (2,000+ words)

Overhyped

  • FAQ schema as a "citation hack"
  • Assuming SEO = AEO success
  • Short, thin content
  • Set-and-forget evergreen content

FAQ

What's the #1 factor for ChatGPT citations?

Referring domains (backlinks) is the strongest predictor. SE Ranking's study of 129,000 domains found sites with 32K+ referring domains are 3.5x more likely to be cited by ChatGPT than sites with fewer backlinks.

Does FAQ schema help with AI citations?

Contrary to popular belief, no. SE Ranking found pages WITHOUT FAQ schema received more citations (4.2 average) than pages WITH FAQ schema (3.6 average). FAQ pages often appear on simpler support pages that naturally earn fewer citations.

How fresh does content need to be for AI citations?

Very fresh. 95% of ChatGPT citations come from content published in the last 10 months. Pages with a visible "last updated" timestamp receive 1.8x more citations than pages without one.

Sources

  1. SE Ranking, How to optimize for ChatGPT: 129,000 domains / 216,524 pages across 20 niches; referring domains, content length, FCP and FAQ-schema findings
  2. Search Engine Journal: coverage of the SE Ranking citation-factor data
  3. Digital Bloom, 2025 AI Citation & LLM Visibility Report: overlap between AI citations and Google rankings
  4. AirOps, The Impact of Stale Content on AI Visibility: content age and citation decay
  5. Moz, Only 12% of AI Mode Citations Match URLs in the Organic SERP (February 2026; nearly 40,000 queries), independent corroboration that most AI citations sit outside the organic top 10 for the same query

See How Your Content Performs

Get a free assessment of your AI citation potential based on the factors that actually matter, or see how our AEO services turn these signals into citations.

Comparing agencies? See the ranking of the 9 best AI SEO and AEO agencies in Singapore (2026): comparison table, methodology, and detailed top-5 profiles.

View ranking