Google’s Discover Algorithm: Understanding Supersession-Decay Filtering for Content Success
Google's Discover Algorithm Now Quietly Filters Your Content Before It Can Rank
Google has been running a two-filter AI system called Supersession-Decay Filtering inside Google Discover for two years — and most publishers never knew it existed.
The stakes are significant. Google Discover serves hundreds of millions of daily active users and billions of monthly active users worldwide. For publishers relying on Discover traffic, understanding this system is no longer optional — it is essential.
What Supersession-Decay Filtering Actually Does
Google published new research detailing how Supersession-Decay Filtering (SDF) works inside Google Discover, the personalized content feed embedded in the Google Search app. The system was quietly deployed two years ago and has been shaping what content reaches users ever since.
SDF operates as a proactive filtering layer that removes stale content before it enters the computation-heavy ranking stage. In plain terms, if your article is deemed outdated, it never gets a chance to compete — regardless of how well it might perform on engagement signals or trend-aware ranking systems.
The research paper, titled Decomposing Staleness in Recommender Systems: A Dual-Filter Framework for Supersession and Decay, explains the core mechanics clearly. "SDF acts as a proactive filtering layer," the researchers wrote. "It prunes fundamentally stale content before it enters the computation-heavy ranking stage, providing a refined candidate set."
The Two Filters That Determine Your Content's Fate
The system breaks staleness into two distinct problems, each handled by a separate AI-powered filter.
The first filter targets supersession — content that has been made obsolete by newer information. The AI evaluates the relationship between older and newer content to determine whether the earlier article still holds value for users. This is not a simple recency check; the system is assessing meaning and relevance, not just timestamps.
The second filter addresses relevance decay — the gradual loss of usefulness that can occur even when no newer content has directly replaced an older piece. An article about a product that has simply aged out of relevance, without ever being contradicted, is a clear example of this quiet erosion.
If either filter flags content as stale, that content becomes ineligible for ranking in Google Discover entirely. Only content that clears both filters moves forward into the ranking pipeline. This is a hard gate, not a soft penalty — and that distinction matters enormously for publishers.
Why Traditional Freshness Signals Were Not Enough
Google's decision to build SDF reflects a well-documented limitation of older freshness detection methods. Two conventional approaches — publication date cutoffs and user engagement metrics — both fail in meaningful ways.
The Problem With Publication Dates
Publication dates measure only when something was written. They cannot determine whether content remains accurate, useful, or relevant to current user needs. Evergreen content published years ago can still answer a query perfectly. Meanwhile, a breaking news article can become misleading within hours of publication. Recency and relevance are not the same thing, and older systems consistently confused the two.
The Problem With Engagement Metrics
Engagement-based systems are inherently reactive. By the time signals indicate that users have stopped clicking or reading a piece of content, those users have already experienced stale recommendations. The damage to user satisfaction happens before the system can correct course.
SDF addresses both failures by predicting staleness before users encounter it. The results validate the approach. Over two years of deployment, user complaints about stale content declined by 54.9%. User engagement showed measurable gains and content diversity improved. The system also reduced computing costs by eliminating outdated content earlier in the pipeline — a notable efficiency benefit at Google's scale.
For publishers, this two-year deployment window is a sobering reminder. The algorithm has been operating quietly while many content strategies remained unchanged. If you have been working to increase traffic to your website through SEO and content, understanding where and why content drops out of the Discover pipeline is now a foundational part of that work.
What Publishers and SEOs Should Do Now
Roger Montti, writing for Search Engine Journal, framed the practical implications clearly. The existence of SDF makes it imperative for publishers to think carefully about their content mix and how long each piece is likely to remain relevant.
Understanding Your Content's Relevance Lifespan
Montti outlined three categories worth examining:
- Evergreen content sustains consistent traffic over time and is less vulnerable to SDF filtering because its relevance does not decay quickly
- Time-limited content — articles tied to trends, events, or developing stories — has a defined window of relevance and may require proactive updates to survive SDF scrutiny
- Breaking news carries a short relevance window by nature and should be planned and resourced accordingly
Mapping your existing content library against these three categories is not a minor housekeeping task. It is a strategic audit that can reveal significant gaps between the content you are producing and the content Google's filters are willing to surface.
The Internal Linking Dimension Publishers Are Missing
One underappreciated implication of SDF involves internal linking strategies. Many content management plugins and AI-based interlinking tools build link structures around semantic relevance alone. SDF adds a freshness dimension to that calculation. Linking an older article to newer related content may help signal that the older piece remains part of an active and updated content ecosystem — rather than an isolated, aging document.
For SEOs, the key question is whether current interlinking and content refresh workflows account for how quickly individual articles lose relevance. A piece that remains semantically related to a topic may still be flagged for relevance decay if nothing connects it to more current coverage. Semantic relevance and temporal relevance are now both in play, and ignoring one while optimizing for the other is a meaningful strategic blind spot. As SEO continues to evolve with AI and algorithm changes, building freshness awareness into linking workflows is becoming a standard requirement rather than an advanced tactic.
Three Practical Steps for Adapting to SDF
Understanding SDF gives publishers and SEOs a concrete framework for auditing existing content libraries.
First, conduct a content audit segmented by expected relevance lifespan. Identify which articles are genuinely evergreen, which are time-sensitive, and which have already likely decayed past the point of Discover eligibility. Prioritize by traffic value so that refresh resources are directed where they will have the most impact.
Second, build a refresh schedule for high-value time-limited content. Updating articles with new data, updated context, or links to more recent coverage can reset relevance signals and potentially restore Discover eligibility. A structured refresh calendar — rather than ad hoc updates — makes this sustainable at scale.
Third, revisit internal linking strategies with freshness in mind. Connecting older evergreen content to newer related articles may help signal ongoing relevance to Google's filtering systems. This is a structural change that does not require rewriting content but can have meaningful traffic implications across a large content library. Pairing this approach with a broader understanding of Google's tools for growing your business online gives publishers a more complete picture of how Google's ecosystem rewards well-maintained, current content.
Rethinking Content Production Through an SDF Lens
Beyond auditing existing content, SDF has implications for how new content should be planned and produced. Publishers who treat every article as a long-term asset without considering its natural relevance lifespan are building a library that will accumulate filtering risk over time.
Assigning a relevance horizon to each piece of content at the time of production — even a rough estimate — creates the foundation for a more defensible Discover strategy. A piece planned with a six-month relevance window should be resourced differently, and refreshed on a different schedule, than a piece designed to remain accurate for three years.
This is not a minor refinement to existing workflows. For publishers heavily dependent on Discover traffic, it represents a meaningful shift in how content is commissioned, managed, and retired. For further context on how Google uses dedicated search and discovery tools to surface content to users, Google's own Discover documentation outlines what signals matter most at the eligibility stage.
Google Discover remains one of the largest passive traffic sources on the open web. Publishers who adapt their strategies to account for SDF are better positioned to maintain visibility in a feed that now filters for freshness before a single ranking signal is applied. The system has been running for two years. The window for catching up is now.
