Google Discover research explains its stale content filtering system

Google has published research that explains how Google Discover filters out stale content before ranking recommendations, and the system is already in use. According to Search Engine Journal, the system is called Supersession-Decay Filtering, or SDF, and it has been deployed for two years.

The research describes SDF as a proactive layer that removes content judged to be stale before it reaches downstream ranking systems. Those later systems still use engagement signals and trend-aware methods, but only after SDF has narrowed the pool of eligible content.

What changed

Google shared new details about SDF, a two-part system used in Google Discover. One filter looks for supersession, which means a page has been overtaken by newer information. The other looks for relevance decay, which means content may lose usefulness over time even if no newer page directly replaces it.

This matters because older ways of handling freshness, such as publication date cutoffs and user engagement drops, do not always reflect whether a page is still useful. A newer date does not guarantee relevance, and an older page can still be valuable if it remains helpful. Google said SDF makes content ineligible for ranking in Discover if either filter identifies it as stale.

What it means for your website

If your site depends on visibility in Google Discover, freshness is not only about publishing often. It is also about whether a page still answers the need it originally served, and whether a newer page has effectively replaced it. A page can lose eligibility in Discover before poor engagement signals show up.

For business owners, this means content planning may need more structure. Evergreen pages may continue to qualify if they stay useful, while time-sensitive pages may need updates, replacement, or a shorter life cycle. The source also notes that freshness may affect interlinking decisions, not just topical similarity.

What to do now

• Review content that covers changing topics, and check whether newer information has made older pages less useful.
• Separate evergreen pages from short-life content, so each type has a different update plan.
• Update or replace pages that have a limited window of relevance, especially if a newer page now serves the same purpose better.
• Look at internal linking on related articles and resources, and consider whether links still point users toward the most current useful page.
• Monitor Discover performance and user feedback patterns for signs that older content is no longer staying eligible or useful.