Content Freshness and AI Citation: Why 50% of AI-Cited Pages Are Under 13 Weeks Old

Content freshness is the recency signal that tells AI search engines whether a piece of content reflects the current state of knowledge. In 2026, it has become one of the five primary signals determining whether an AI engine cites your page - alongside relevance, authority, structure, and retrievability. Roughly half of all pages cited by ChatGPT, Perplexity, and Gemini are under 13 weeks old, and content published within the past 30 days earns an estimated 3.2 times more citations than equivalent content published more than a year ago.

Why AI engines weight freshness differently from Google

Google has incorporated freshness signals for years through its Query Deserves Freshness (QDF) algorithm, which boosts recency for news-like topics. AI engines apply freshness more broadly - and more strictly. The reason is structural: when an AI model synthesises an answer, it is making an implicit claim about the current state of affairs. Citing a two-year-old article about a fast-moving topic risks producing an inaccurate answer, which damages user trust in the AI product itself.

The result is a consistent preference for recently published or recently updated content across all major AI answer engines. This does not mean old content cannot be cited - cornerstone definitions and methodological guides have a longer shelf life - but a page published in 2023 and never touched since is at a structural disadvantage compared to a competitor who covered the same topic last month.

The 13-week rule explained

Research from multiple sources in early 2026 converged on a striking finding: approximately 50% of AI-cited content is less than 13 weeks old at the time of citation. This 13-week rule is not a published algorithm parameter - it is an observed pattern in citation behaviour across ChatGPT, Perplexity, and Gemini, sampled across thousands of queries.

The practical implication is a decay curve. A well-optimised page earns maximum citation exposure in the first 13 weeks after publication or meaningful update. After that, citation rates decline gradually - dropping by roughly 50% within 12 months for content on evolving topics such as tooling comparisons, statistics, policy guidance, and pricing. Evergreen definitional content decays more slowly, but still decays.

What counts as a meaningful update

AI engines do not respond to superficial date changes. Updating the last-modified timestamp on a page without changing body content is detectable - crawlers compare successive versions and can identify whether substantive content changed. A meaningful update includes at least one of the following:

• Adding a new section that addresses a development in the topic area since original publication
• Updating statistics, data points, or tool references that have changed
• Expanding the FAQ section with questions that reflect current user intent
• Revising claims that are no longer accurate due to product, platform, or policy changes
• Adding an explicit "Updated June 2026" marker near the top of the page with a short summary of what changed

What does not fool AI engines

Updating the isoDate in your schema markup without corresponding body changes has no effect on citation rate. AI systems using retrieval-augmented generation pull live content from the page, not just the schema date. They compare the substance of the content against the implied recency of the date. A page that claims to be from 2026 but references outdated products as cutting-edge will be treated as stale regardless of its markup.

The five freshness signals AI engines check

Content freshness is not a single signal but a cluster of five detectable indicators. Optimising all five compounds the effect:

• Schema dateModified - the machine-readable last-updated date in Article or BlogPosting schema, set to the date of the most recent genuine content change
• Visible publication and update dates - an explicit "Published: January 2026, Updated: June 2026" note near the top of the article, readable by both users and crawlers
• Temporal qualifiers in the body - phrases like "as of June 2026" anchored near time-sensitive claims to give AI engines a specific recency reference
• Recency of referenced sources - outbound links to sources published within the past 12 months signal that the content reflects current knowledge
• Crawl frequency - pages that are crawled regularly because they are linked internally, shared externally, or included in a freshly submitted sitemap are treated as actively maintained

Building a content refresh programme

A systematic refresh programme is the highest-ROI AEO activity for sites with an existing content library. The goal is to cycle your most citation-eligible pages through meaningful updates on a rolling basis, keeping as many pages as possible inside the 13-week freshness window at any given time.

Step 1: Audit your content by topic volatility

Divide your posts into three tiers. Evergreen content - definitions, foundational guides, methodology posts - decays slowly and needs refreshing every 9 to 12 months. Semi-volatile content - tool comparisons, strategy posts, benchmark studies - needs refreshing every 4 to 6 months. Volatile content - statistics posts, best-of lists, platform-specific guides - needs refreshing every 8 to 13 weeks.

Step 2: Prioritise by AEO citation surface

Not all content is equally worth refreshing. Prioritise pages that already have strong structured data, a logical heading hierarchy, and existing FAQ sections. These pages are more citable to begin with - refreshing them returns more citation value per hour of effort than refreshing pages with weak AEO fundamentals.

Step 3: Write updates that add real substance

For each refresh, aim to add 150 to 300 words of genuinely new content. The most reliable method is to mine your support inbox, sales call transcripts, and recent search queries for questions that emerged after original publication. Answering those questions as new FAQ entries or a short new section satisfies both the freshness signal and adds new citation surface area.

Freshness signals in Schema markup

The dateModified field in Article or BlogPosting schema is the primary machine-readable freshness signal. Set it to the ISO 8601 date of the last meaningful content update - not the date you corrected a typo or swapped an image. Keep datePublished set to the original publication date and leave it unchanged. AI models cross-reference both dates to assess how actively maintained the content is.

{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "headline": "Your article title",
  "datePublished": "2026-01-15",
  "dateModified": "2026-06-18",
  "author": {
    "@type": "Person",
    "name": "Jane Smith"
  }
}

Evergreen content and a different freshness strategy

Not all content ages at the same rate. A page defining what schema markup is, or explaining how retrieval-augmented generation works, is inherently more stable than a page comparing specific tool pricing. For evergreen content, the freshness strategy shifts: focus on depth and authority signals rather than frequent updates. A well-structured, comprehensive guide stable for 12 months may still outperform a superficially refreshed competitor if its schema, E-E-A-T signals, and citation surface are stronger.