Do publication dates and statistics improve AI search citation rates?

Yes: freshness metadata and statistics are the two strongest verifiable signals for AI citation in the data.

A cross-platform arXiv study tested citation predictors across Brave, Google, and Perplexity simultaneously. Freshness metadata (datePublished and dateModified signals) showed r=0.68 correlation with citation rates, above semantic HTML (r=0.65) and structured data (r=0.63). That is the strongest measured on-page signal across all three platforms.

Statistics separately deliver +37% AI visibility improvement per an arXiv GEO study, second only to adding citations (+40%). The two signals compound: a dated, statistics-rich page is both fresher and more citable than a page with either signal alone.

Why dates matter: the freshness bias in AI retrieval

AI search is dramatically fresher in its citation preferences than traditional search. A separate arXiv study found median AI-cited content age is 148 days, versus 493 days for equivalent Google results in the automotive vertical. An ALM Corp analysis found 30-day-old content receives 3.2× more citations than content over 90 days old.

83.3% of AI-cited sites published 2025 content, versus 23.3% of a random web sample, a 257% over-representation. Perplexity is the most extreme: approximately 50% of its citations link to 2025 content alone.

The practical implication: visible publication dates and dateModified schema make your freshness legible to AI retrieval systems. An undated page is disadvantaged even if the content is current, because AI systems cannot confirm when it was published.

Why statistics matter: verifiable claims as citation targets

AI systems retrieve sources to attribute. A content claim with a specific statistic: "47% of marketing teams have adopted AI tools in 2025 (Salesforce State of Marketing)": is a citable unit: the AI can include the claim and the source. A qualitative claim: "many marketing teams are adopting AI tools": cannot be cited directly.

Statistics also serve as entity anchors: named statistics associated with named studies create traceable attribution chains that AI systems can follow. An AI composing a response about marketing automation will retrieve the page containing the named statistic, not the page describing the general trend without data.

The format that maximises both signals

A page with a visible publication date, clearly formatted statistics, Article schema with datePublished and dateModified, and inline source attribution combines all four signals simultaneously. The date confirms freshness, the statistics provide citable units, the schema makes both machine-readable, and the attribution connects the content to the credibility of the cited sources.

What the evidence doesn't prove

The r=0.68 freshness correlation is cross-platform but correlational. Pages with current freshness metadata are also more likely to be actively maintained, which correlates with other quality signals that independently predict citation rates.

The +37% figure for statistics is an uplift from adding statistics to existing content. The improvement assumes baseline content without statistics; content already statistics-rich would see a smaller marginal gain from adding more.