GEO Glossary

/terms/freshness-signals · 3 min read · intermediate

Freshness signals

Freshness signals are the metadata inputs AI engines and search engines use to assess how recent a piece of content is — datePublished, dateModified, version history, and on-page recency markers. Freshness weight has increased meaningfully in the AI-search era because RAG retrieval favors recently-updated sources.

Citation status

ChatGPT—Perplexity—Claude—Copilot—Gemini—

Last checked 2026-05-14

What are freshness signals?

Freshness signals are the metadata and on-page inputs engines combine to decide how recent a piece of content is. The primary signals:

  • datePublished — when the page was first published (ISO date in Article schema)1.
  • dateModified — when the page was last meaningfully updated.
  • On-page recency markers — "Last updated" labels, year references in headings, version numbers.
  • Crawl-history signals — how often engines have observed the page changing.
  • External recency signals — when other authoritative pages have linked to or cited this page recently.

Engines combine these into a freshness score that influences retrieval and ranking for recency-sensitive queries.

Status in 2026

Increasingly weighted, increasingly audited. AI engines (especially Perplexity and ChatGPT search) weight freshness heavily on time-sensitive queries — vendor pricing, product launches, recent announcements. Google's classical search and AI Overview both inherit freshness signals from Article schema and on-page markers. The 2024–2026 trend: engines have gotten better at detecting date spoofing (bumping dateModified on unchanged content) and increasingly penalize sites that do it.

How to apply

Freshness is signal-quality, not signal-quantity. Three concrete moves:

  • Update content when it actually changes: every legitimate revision bumps dateModified. Calendar-driven "freshness updates" that don't change substance are increasingly detectable by content-diff inspection.
  • Match on-page freshness markers to schema dates: if you ship <time datetime="2026-05-14">Last updated May 14, 2026</time> on the page, the schema's dateModified should match. Mismatches trigger ambiguity flags.
  • Track freshness decay per content type: definitions and core concepts can hold for years without revisiting; vendor comparisons and product mentions tend to need updates every 3–6 months. Tag each page with its expected revision cadence and audit on schedule.

What to skip: bulk date-bumps on unchanged content. Engines detect this via content-diff (in Google's case) or re-embedding-vs-stored-embedding comparison (in AI engines' case). The penalty tends to be suppression of freshness credit, not just neutrality.

How it relates to other concepts

  • Direct input to authority signals — freshness is one of the four authority layers.
  • Component of E-E-A-T — Trustworthiness includes accurate freshness representation.
  • Exposed via Article schema datePublished and dateModified properties.
  • Critical factor in RAG retrieval — many production RAG systems run recency-filtered retrieval as a default.

Footnotes

  1. Google Search Central: Article structured data implementation guide covering datePublished and dateModified. developers.google.com/search/docs/appearance/structured-data/article. ↩

Related terms

FAQ

Does bumping dateModified actually help?
Yes — if the underlying content genuinely changed. Engines increasingly cross-check `dateModified` against content-diff to detect spoofing. Legitimate updates earn freshness credit; date-only updates on unchanged content increasingly trigger penalties.
How often should I update content?
Update when there's a real change — new data, new vendor pricing, corrected claim, evolved best practice. Calendar-driven updates without real content change tend to underperform calendar-driven updates that actually rewrite substance.
Do AI engines weight freshness differently from Google's classical search?
AI engines tend to weight freshness more heavily because RAG retrieval often runs against recency-filtered indices. For time-sensitive queries (vendor pricing, recent product launches), stale content is sometimes filtered out rather than just deprioritized.

Sources & further reading