Knowledge base/ExtractabilityAUTO

Content updated within 3 months

There is a measurable freshness cliff at three months. Older content gets cited less.

The 3-month freshness cliff

AI assistants strongly prefer recent content. The falloff is not gradual — research shows a sharp drop in citation frequency for content older than about three months, especially in fast-moving categories like software, news, and industry trends.

This is not because old content is wrong. It is because models cannot tell whether old content is still accurate, and they hedge against that uncertainty by preferring recent sources when they are available.

How to signal freshness

  • Publish a visible last-updated date on every article or documentation page.
  • Use schema.org Article markup with dateModified.
  • Update sitemap lastmod dates when you change content.
  • Do actual refreshes, not fake ones. Update the facts, not just the date.
  • Retire content that is no longer accurate rather than letting it linger.