Does content freshness affect AI search citation rates?

Yes: the freshness bias in AI search is substantially stronger than in traditional search.

An arXiv study comparing AI and traditional search citation ages found the median article age in Claude's citations was 148 days, versus 493 days for equivalent Google search results in the automotive vertical. In consumer electronics, AI median citation age was 62–90 days versus 130 days for Google. AI systems are not just slightly fresher in their citation preferences: they are dramatically fresher.

An ALM Corp analysis of 1.2 million ChatGPT responses quantified the penalty directly: 30-day-old content received 3.2× more citations than content over 90 days old, controlling for other variables.

The freshness curve peaks at 30–89 days, not 0–30

That framing can mislead. An AirOps study found the peak citation rate is not for the very freshest content. Pages aged 30–89 days achieved the highest citation rate at 32.8%. Content under 30 days old hit only 25.3%. The likely explanation: very new content has not yet been indexed, processed, and incorporated into retrieval pools.

The practical implication for content teams: the highest return on freshness investment comes from updating existing pages at the 30–60 day mark, not from publishing new pages as frequently as possible.

Freshness metadata is the strongest on-page citation signal

An arXiv study testing cross-platform citation predictors across Brave, Google, and Perplexity found metadata and freshness signals showed the strongest correlation with citation rates at r=0.68: above semantic HTML (r=0.65) and structured data (r=0.63). The same study found datePublished and dateModified schema are the primary mechanisms through which freshness is assessed at retrieval time.

The Amsive figure from the introduction: 50% of AI-cited content is under 13 weeks old: is consistent with a separate Savannabay analysis that found 83.3% of AI-cited sites published 2025 content, versus 23.3% of a random web sample. That's a 257% over-representation of recently published sites relative to the web at large.

Perplexity vs ChatGPT freshness profiles

The freshness bias is not uniform across platforms. A Seer Interactive analysis of 5,000+ URLs found approximately 50% of Perplexity's citations are from 2025 content alone: approximately 80% from 2023–2025. Perplexity is a live-web retrieval system with explicit recency weighting. ChatGPT's freshness profile is less extreme but still significantly fresher than traditional search.

What the evidence doesn't prove

The ALM Corp finding (3.2× more citations for 30-day content) is not a controlled experiment. Content published in the last 30 days is also more likely to cover fast-moving topics, which may receive freshness weighting specifically because their relevance decays quickly. The freshness effect may be stronger for time-sensitive categories than for evergreen content.

The arXiv study (r=0.68 for freshness metadata) is a correlation. It cannot distinguish between freshness signals causing citations and high-quality, actively maintained pages being both more likely to have proper metadata and more likely to be cited.

Both findings agree directionally: newer is better, and legible freshness signals (visible dates, dateModified schema) matter more than age alone.