What Drives ChatGPT Citations? New Data Reveals Key Factors

Understanding what drives content citations within AI systems like ChatGPT has become a crucial focus for digital marketers and content creators. A recent large-scale analysis examined over 129,000 unique domains across more than 216,000 pages spanning twenty distinct niches to identify the core factors that correlate with how frequently ChatGPT references specific sources. The findings reveal that the number of referring domains stands out as the single strongest predictor of whether a site earns citations, though several other elements also play significant roles.

Backlinks and Domain Authority

Link diversity demonstrated the clearest relationship with citation frequency. Websites supported by up to 2,500 referring domains typically received between 1.6 and 1.8 citations on average. In stark contrast, sites with more than 350,000 referring domains averaged 8.4 citations. Researchers observed a notable threshold effect around 32,000 referring domains, where citation numbers nearly doubled from 2.9 to 5.6.

Domain Trust scores followed a comparable trajectory. Sites scoring below 43 averaged just 1.6 citations, while those in the 91-96 range jumped to 6 citations. The most trusted domains, scoring between 97 and 100, achieved the highest average of 8.4 citations. Interestingly, Page Trust appeared less influential than domain-level authority signals. Any page scoring 28 or above for Page Trust received approximately 8.3 citations, suggesting ChatGPT prioritizes overall domain credibility over individual page metrics.

Contrary to common assumptions, .gov and .edu domains didn’t automatically outperform commercial websites. Government and educational domains averaged 3.2 citations compared to 4.0 for sites without these trusted designations. The study authors emphasized that “what ultimately matters is not the domain name itself, but the quality of the content and the value it provides.”

Traffic Volume and Search Performance

Domain traffic emerged as the second most significant factor, though its impact only became apparent at higher volumes. Sites receiving under 190,000 monthly visitors averaged between 2 and 2.9 citations regardless of their exact traffic numbers. A domain with just 20 organic visitors performed similarly to one with 20,000. The correlation strengthened dramatically after crossing the 190,000 monthly visitor threshold, with domains exceeding 10 million visitors averaging 8.5 citations.

Homepage traffic specifically influenced citation rates. Sites attracting at least 7,900 organic visitors to their main page demonstrated the highest citation frequencies. Average Google ranking positions also aligned with ChatGPT citations. Pages ranking between positions 1 and 45 averaged 5 citations, while those ranking between 64 and 75 averaged only 3.1. The researchers noted that while this doesn’t confirm ChatGPT uses Google’s index, it suggests both systems evaluate authority and content quality through similar lenses.

Content Quality and Structure

Content length showed consistent correlation with citation likelihood. Articles shorter than 800 words averaged 3.2 citations, while those exceeding 2,900 words averaged 5.1 citations. Structure proved important beyond mere word count. Pages with section lengths between 120 and 180 words separating headings performed best, averaging 4.6 citations. Extremely brief sections under 50 words averaged only 2.7 citations.

Including expert quotes boosted citation rates to 4.1 compared to 2.4 for content without such endorsements. Articles containing 19 or more statistical data points averaged 5.4 citations versus 2.8 for pages with minimal data. Content freshness produced one of the study’s clearer findings: pages updated within three months averaged 6 citations, while outdated content averaged just 3.6.

Surprisingly, pages featuring FAQ sections actually received fewer citations (3.8) than those without them (4.1). The researchers noted their predictive model viewed FAQ absence as a negative signal, suggesting this discrepancy might occur because FAQs often appear on simpler support pages that naturally earn fewer citations. Question-style headings also underperformed straightforward headings, earning 3.4 citations versus 4.3, contradicting conventional voice search optimization advice.

Social Engagement and Review Platforms

Brand mentions on discussion platforms showed strong correlation with citation frequency. Domains with minimal Quora presence (up to 33 mentions) averaged 1.7 citations, while those with heavy Quora representation (6.6 million mentions) corresponded to 7.0 citations. Reddit demonstrated similar patterns, with domains receiving over 10 million mentions averaging 7 citations compared to 1.8 for those with minimal activity.

The authors highlighted this finding as particularly relevant for smaller websites: “For smaller, less-established websites, engaging on Quora and Reddit offers a way to build authority and earn trust from ChatGPT, similar to what larger domains achieve through backlinks and high traffic.”

Presence on review platforms including Trustpilot, G2, Capterra, Sitejabber, and Yelp also correlated with increased citations. Domains listed on multiple review platforms earned between 4.6 and 6.3 citations on average, while those absent from such platforms averaged only 1.8.

Technical Performance Factors

Page speed metrics consistently correlated with citation likelihood. Pages with First Contentful Paint under 0.4 seconds averaged 6.7 citations, while slower pages exceeding 1.13 seconds averaged just 2.1. Speed Index showed similar patterns, with sites scoring below 1.14 seconds performing reliably well and those above 2.2 seconds experiencing steep declines.

One counterintuitive finding emerged regarding Interaction to Next Paint scores. Pages with the fastest INP scores (under 0.4 seconds) actually received fewer citations (1.6 average) than those with moderate scores between 0.8 and 1.0 seconds (averaging 4.5 citations). Researchers suggested extremely simple or static pages might not signal the depth ChatGPT seeks in authoritative sources.

URL and Title Construction

The analysis found that broad, topic-describing URLs outperformed heavily keyword-optimized ones. Pages with low semantic relevance between their URL and target keyword (scoring 0.00 to 0.57) averaged 6.4 citations, while those with highest semantic relevance (0.84 to 1.00) averaged only 2.7. Titles followed the same pattern, with low keyword matching averaging 5.9 citations versus 2.8 for highly optimized titles. The researchers concluded that “ChatGPT prefers URLs that clearly describe the overall topic rather than those strictly optimized for a single keyword.”

Underperforming Factors

Several commonly recommended AI optimization tactics showed minimal or negative correlation with citations. FAQ schema markup underperformed, with pages using FAQ schema averaging 3.6 citations compared to 4.2 for those without. LLMs.txt files demonstrated negligible impact, and outbound links to high-authority sites showed minimal effect on citation likelihood.

Strategic Implications

These findings suggest that established SEO strategies may already support AI visibility objectives. If you’re building referring domains, generating traffic, maintaining fast-loading pages, and keeping content updated, you’re addressing the factors this report identified as most predictive. For smaller sites without extensive backlink profiles, the research points to community engagement on platforms like Reddit and Quora as viable paths to building authority signals. The data also recommends focusing on content depth rather than keyword density.

The researchers emphasized that these factors operate interdependently. Optimizing one signal while neglecting others reduces overall effectiveness. It’s worth noting that this analysis specifically examined ChatGPT, and other AI systems may weight factors differently. Since the study doesn’t specify which ChatGPT version or timeframe the data represents, these patterns should be treated as directional correlations rather than definitive proof of how ChatGPT’s ranking algorithm functions.

(Source: Search Engine Journal)