Google Discover has implemented a system to remove stale articles before they are ranked, significantly altering content visibility for publishers, the company announced.
The system, known as Decomposing Staleness in Recommender Systems (SDF), has operated in Discover for two years, resulting in a 54.9 percent reduction in reader reports of outdated content, according to Google.
SDF identifies articles as stale through two primary mechanisms: when newer, more relevant content replaces them, or when they have reached most of their expected audience and are considered ‘worn out.’
The staleness assessment occurs prior to the ranking process, meaning articles flagged as stale are not presented to users, regardless of potential engagement metrics.
Publishers must adjust their content strategies to reduce negative impacts, Google advised. Key recommendations include publishing timely updates to existing content and maintaining precise angles for articles.
Additionally, publishers should avoid using dated language in content intended to be timeless and develop a clear understanding of Discover’s specific traffic metrics.
Google also cautioned against republishing content as a strategy to circumvent the staleness filter, noting that such practices are ineffective.
The scientific paper detailing the SDF system is scheduled for presentation at the CIKM conference in Rome from Nov. 7-11.
Source: Abondance
