Who does AI cite? The 5 factors from research

Whether an AI cites your website depends on measurable factors. The Princeton GEO study (KDD 2024, 10,000 queries across 9 domains) and Vercel’s analysis of 500+ million AI crawler fetches point to five of them. Our AI visibility check measures exactly these five.

1. Be readable without JavaScript

The most important and most overlooked factor: AI crawlers do not execute JavaScript. Vercel analyzed 500+ million GPTBot fetches and found zero evidence of JavaScript execution. The same holds for ClaudeBot and PerplexityBot. A site that builds its content in the browser (client-side rendering) can rank #1 on Google and still be completely blank to every AI. So check your raw HTML. Whatever is not in it, the AI never sees.

2. Numbers & statistics: +41% citation visibility

The Princeton researchers tested nine optimization tactics. The strongest: adding concrete statistics. It raised visibility in AI answers by up to 41%. Prices, quantities, percentages, years: AI systems prefer to cite passages that carry verifiable facts. Pages around position 5 gained the most (+115%), so GEO helps most where a site is not already on top.

3. Cite sources & use quotations

Right behind statistics, the study ranks “cite sources” and “quotation addition”: content that backs up claims with external sources or direct quotes gets picked up significantly more often. A plausible reason: for answer engines, sourced content is the lower risk.

4. Answer structure: FAQ, lists, tables

AI search runs on retrieval-augmented generation (RAG): pages are split into chunks, and the chunk that answers the user question most directly gets cited. Questions as headings (plus FAQPage schema), lists and tables make your content chunkable. ChatGPT extracts a lot from few sources (~7 per answer), Perplexity cites broadly (~16 sources): clear structure helps with both.

5. Freshness: you lose ground after 60 to 90 days

Perplexity explicitly weighs recency: content without an update signal measurably loses citation probability after 60 to 90 days. A machine-readable date (dateModified in JSON-LD, <time datetime>) and regular maintenance of key pages therefore directly affect your chance of being cited.

What to do now

1. Have your site rated by the AI visibility check. Its “Citability (GEO)” category tests all five factors automatically.
2. Make sure no crawler is locked out: AI crawler check.
3. Give AI your facts first-hand: create your llms.txt and get listed.

Sources: Aggarwal et al., “GEO: Generative Engine Optimization” (Princeton/Georgia Tech/Allen AI/IIT Delhi, KDD 2024) · Vercel: The rise of the AI crawler (500M+ fetch analysis).