Citation and Evidence Density

Citation and Evidence Density measures whether your page backs its claims with verifiable evidence — outbound links, statistics, quotes, data tables, and a references section — so AI engines can treat it as a citable source. To pass, a page needs 3 or more outbound external links AND at least one statistical marker. It sits in the Authority and Trust group of the friendly4AI v4.5 methodology.

What does this parameter check?

friendly4AI parses your page DOM and counts five evidence signals:

• Outbound links to external domains. Same-domain links and # anchors don't count; only http/https hosts do.
• Statistical markers — percentages, large comma-grouped numbers, and quantified figures such as "3.2 million" or "45%".
• Quotes — <blockquote> elements plus attribution phrases like "according to" or "said".
• Data tables — <table> elements.
• References section — flagged by labels such as "references", "sources", "bibliography", or "citations".

From those signals the scanner derives an internal evidence_density_score and an overall band.

Why does evidence density matter for AI-readiness?

AI engines would rather cite content that is itself well-sourced. Links to credible domains, quantified data, and a references section tell a model that your claims are checkable instead of merely asserted — the same markers a human fact-checker hunts for. Once a model can trace your statements back to external evidence, it treats the page as a citable secondary source rather than an unverifiable opinion. A thin, unsourced page rarely gets quoted in an AI answer, even when its underlying claim happens to be correct. Reinforce the signal by pairing this with author authority and a strong dated-statistics ratio.

How is it scored?

The v4.5 methodology scores this parameter in three tiers, all drawn from the evidence signals above:

• Pass (100) — the page has 3 or more outbound external links AND at least one statistical marker. That combination shows genuine sourcing.
• Partial (50) — a single signal is present (one outbound link, statistic, quote, or data table), but the full pass combination is missing.
• Fail (0) — none of those four signals appear.

The processor also reports an evidence_density_score for finer diagnostics, weighting the signals: up to 40 points for outbound links, 25 for statistics, 10 each for quotes and tables, and 15 for a references section. That diagnostic number doesn't move the band. The pass/partial/fail result still comes down to the outbound-link-plus-statistic rule above.

How do I fix a low score?

• Add at least 3 outbound links to credible external sources — research, official docs, primary data — on substantive pages. Same-domain links won't help here.
• Back claims with quantified, dated statistics instead of vague phrasing, so the statistical-marker signal fires alongside the outbound links and pushes you to a pass.
• Wrap quoted material in <blockquote> and attribute it ("according to ...") so the quote signal registers.
• Put comparative data in real <table> elements, and add a "References" or "Sources" section to longer pieces.
• Re-scan, then check the outbound_link_count, statistic_markers, quote_count, data_table_count, and reference_section evidence fields.