Citation Probability: What Makes AI Systems Choose Your Content as a Source
Citation Probability is not a proxy for backlinks. It is a seven-factor model of the signals AI systems use when deciding which content to cite in a generated response. High-authority domains with low factual density score worse than lower-authority domains with dense, specific, structured content.
When an AI system generates a response and decides which sources to cite, it is not running a PageRank calculation. It is filtering retrieved chunks through a citation eligibility model that weights specific content properties. A chunk with high factual density, specific verifiable claims, and clear topical authority will be selected over a generic overview from a higher-authority domain. SiteNexis models this with seven weighted factors derived from the observable patterns in how AI systems attribute sources.
The Seven Citation Factors
- Factual Density (20%) — specific, verifiable claims per 100 words. Generic assertions score near zero.
- Claim Specificity (15%) — claims that name entities, dates, percentages, or other verifiable details.
- Primary Entity Authority (15%) — Entity Confidence Score of the page's primary entity.
- Topical Authority Depth (15%) — breadth and depth of supporting content across the domain on this topic.
- Structural Citation Readiness (15%) — structured paragraphs, defined claims, schema markup that makes attribution possible.
- Temporal Freshness (10%) — recency of content, presence of datePublished/dateModified schema.
- Trust Signal Density (10%) — authorship attribution, organisation schema, external validation signals.
Why Factual Density Matters Most
Factual density is the single highest-weight factor because citation is fundamentally about transferring a specific piece of information. A page that states "machine trust is important for AI visibility" has near-zero factual density on that claim. A page that states "domains with Entity Confidence Score below 50 see citation probability decline by approximately 30–40% relative to domains scoring above 75" has high factual density. The AI system can quote the second statement. It has nothing specific to quote from the first.
◆Increase factual density by replacing generic assertions with specific observations. Instead of "entity consistency is important," write "entity consistency failures — name mismatches between schema and body text — are present in over 60% of domains we have audited, and they reduce Entity Confidence Score by an average of 18 points." The specific version is citable. The generic version is not.
Structural Citation Readiness
Structural citation readiness is the least obvious factor. AI systems need content structured so that individual claims can be extracted without losing their meaning. A well-structured citable claim has: a clear subject (the entity or concept the claim is about), a specific predicate (what is being asserted about the subject), and evidence or attribution (a data source, observed pattern, or methodology reference). Claims buried in long compound sentences with multiple qualifications are structurally difficult to cite without distortion.
Topical Authority Depth
A single page with high factual density will not score highly on topical authority depth if it is the only page on the domain covering this topic. AI systems prefer to cite from domains where the topic is covered in depth across multiple pages — this pattern signals genuine expertise rather than isolated content production. The Perception Graph's topical authority clusters feed directly into this factor: topics that appear as well-supported clusters (multiple pages, multiple entity relationships) score higher than isolated topic pages.