Scores documents by inception date and average link-acquisition rate, treating page age and the velocity of incoming links as ranking signals. Foundational age-aware ranking that distinguishes established documents from newly minted ones.
Patent Overview
- Inventor
- Jeffrey Dean, others
- Assignee
- Google LLC
- Filed
- 2003
- Granted
- 2010-11-23
The Challenge
The Challenge
Page age carries information. A document that has existed for years with a steady link rate has been validated by time and the web's link graph. A document published yesterday with a sudden link surge may signal news or may signal manipulation. Scoring needs to read these signals carefully.
- New Documents Are Underranked Without Age Signal — Pure link-based scoring penalizes new high-quality pages until they accumulate links. Age-aware scoring lets newly published content compete on intrinsic signals.
- Old Documents Earn Trust Through Time — Documents that have existed and earned links steadily over years carry an established-authority signal that raw scoring misses.
- Link-Acquisition Velocity Distinguishes Genuine From Manipulated — Sudden link spikes can signal news interest or coordinated manipulation. Velocity analysis distinguishes the patterns.
- Inception Date Must Be Estimated Reliably — Documents don't always carry accurate publication dates. The system must estimate inception from first crawl, first link, and embedded metadata, validating against manipulation.
- Age Cuts Both Ways — Old documents on freshness-sensitive queries are stale; old documents on evergreen queries are authoritative. Age must combine with query freshness sensitivity.
Innovation
How The System Works
The system estimates per-document inception date from first-crawl and first-link signals, tracks per-document link-acquisition velocity over time, distinguishes natural from manipulated link patterns, and combines age and velocity into a tunable scoring factor.
- Estimate Inception Date — Per document, infer inception from first crawl date, earliest inbound link, embedded publication metadata, and content-archive signals. Cross-validate to resist manipulation.
- Track Link Acquisition Over Time — Per document, record the timestamped sequence of inbound link discoveries. Compute per-window acquisition rates.
- Compute Average Link Velocity — Rolling-window averages capture steady-rate, accelerating, and decaying patterns. Velocity is itself a ranking signal.
- Distinguish Natural From Manipulated Spikes — Sudden link spikes correlate with news, viral content, or manipulation. The system applies pattern analysis to discriminate.
- Combine With Query Freshness — On freshness-sensitive queries, recent inception with strong velocity rewards. On evergreen queries, age-with-steady-velocity rewards instead.
- Apply Trust Gating — Domain trust attenuates age and velocity rewards. Low-trust domains can't fake aging or buy link velocity past a cap.
- Cache Scores In Index — Age-and-velocity score caches per document. Rankers consume at query time without recomputing.
Age Plus Velocity
The patent's load-bearing idea is that inception date alone is too coarse. Combined with link-acquisition velocity over time, age becomes a rich signal that distinguishes established authority, viral surge, and manipulation.
Time Is A Ranking Dimension
Documents exist on a timeline. Their position on the timeline, the slope of their link acquisition, and the relationship between the two are all ranking-relevant signals that the index must capture.
- Inception-Date Estimation — First-crawl, first-link, and metadata signals cross-validate to estimate when a document came into existence.
- Velocity-Over-Time — Rolling-window link-acquisition rates capture pattern: steady, accelerating, decaying, or spiking.
- Pattern-Aware Scoring — Natural growth patterns versus manipulated spikes earn different treatment. Pattern analysis is the discriminator.
Technical Foundation
Technical Foundation
The patent specifies the inception estimator, link-history store, velocity calculator, pattern classifier, freshness combiner, and trust gate.
- Inception Estimator — Cross-validates first-crawl, first-link, embedded metadata, and archive signals to estimate document inception. Manipulation-resistant by design.
- Link-History Store — Timestamped record of every inbound link discovery. Enables rolling-window velocity calculation.
- Velocity Calculator — Computes per-document link acquisition rate over multiple windows. Captures pattern (steady, accelerating, decaying, spiking).
- Pattern Classifier — Distinguishes natural growth from manipulated spikes. Classifier-aware ranking applies different treatment to each pattern.
- Freshness Combiner — Combines age and velocity with per-query freshness weight. Output is a per-document, per-query age factor.
- Trust Gate — Per-domain trust attenuates age and velocity rewards. Low-trust domains can't game by faking aging or buying velocity.
The Process
The Process
Inception estimation runs at document discovery; velocity calculation runs continuously. Age-and-velocity scores cache per document for query-time consumption.
- Discover Document — First crawl ingests document. Inception estimator runs cross-signal estimation.
- Initialize Link History — Link-history store opens a per-document timestamped record.
- Continuous Crawl Updates — Each subsequent crawl appends discovered inbound links to history.
- Compute Velocity Windows — Rolling-window calculator updates velocity scores per document.
- Classify Pattern — Pattern classifier tags growth as natural, spiking, decaying, or manipulated.
- Cache Score — Combined age-and-velocity score caches per document. Index update propagates.
- Apply At Query Time — Per query, freshness combiner multiplies base score by query-weighted age factor. Trust gate attenuates.
Quality Control
Quality Control
Inception and velocity signals are targets for manipulation. The patent specifies safeguards.
- Multi-Signal Cross-Validation — Inception estimation cross-validates first-crawl, first-link, metadata, and archive signals. No single signal can be faked into trust.
- Pattern Classifier — Distinguishes natural from manipulated link patterns. Manipulated spikes earn no boost or active demotion.
- Velocity Bounds — Velocity score is bounded. Pay-for-links velocity past a threshold stops accumulating reward.
- Trust Gating — Per-domain trust attenuates rewards. Low-trust domains earn less from age and velocity signals.
- Continuous Recalibration — Classifier and scoring weights recalibrate against fresh labeled data. New manipulation patterns trigger model updates.
Real-World Application
Inception-date and velocity signals are foundational to every modern ranking system. The pattern of natural growth versus spike is the structural discriminator that defines what manipulation looks like in the link graph.
- Cross-validated Inception Estimation — Multi-signal estimation resists single-vector manipulation. First-crawl, first-link, metadata, and archive cross-check.
- Rolling-window Velocity Method — Per-document link-acquisition rate over time. Captures steady growth, acceleration, decay, and spikes.
- Pattern-aware Scoring Discriminator — Natural growth versus manipulated spike earn different treatment. The pattern is the signal.
Why Steady Link Growth Wins
Steady, organic link acquisition earns the highest age-and-velocity rewards. Pattern classifiers flag spikes as suspicious; bursts must align with news or viral signals to earn reward.
Why Faking Page Age Doesn't Work
Multi-signal cross-validation means changing one signal (publication date in metadata) doesn't move the inception estimate. First-crawl, first-link, and archive evidence anchor age estimation.
<\/section>What This Means for SEO
What This Means for SEO
This patent scores documents by their estimated inception date and the velocity of inbound-link acquisition over time, distinguishing time-validated authority from sudden manipulation. SEO implication: earn links steadily and let pages age naturally, because the pattern of accumulation, not the raw count, is what the system reads.
- Steady Link Velocity Beats Bursts — The system computes rolling-window link-acquisition rates and classifies them as steady, accelerating, decaying, or spiking. A consistent trickle of new links signals organic interest, while a sudden spike with no news or viral trigger looks manipulated and earns little or negative reward.
- Do Not Fake Publication Dates — Inception is cross-validated from first crawl, first inbound link, embedded metadata, and archive signals. Changing the date in your CMS or schema does not move the estimate, so backdating content to fake authority is wasted effort.
- Match Effort To Query Freshness — Age combines with per-query freshness sensitivity: recent inception plus strong velocity wins on news-style queries, while age plus steady velocity wins on evergreen ones. Decide whether a topic rewards recency or established authority before you invest.
- Trust Caps Limit Bought Velocity — Velocity reward is bounded and gated by domain trust, so a low-trust site cannot buy its way past the cap. Build domain trust first; link velocity only converts into ranking on a trusted host.
- New Pages Can Compete On Intrinsic Signals — The age signal exists partly so genuinely new high-quality pages are not buried until they accumulate links. Publish strong content confidently rather than gating launches on a pre-built link campaign.
- Spikes Need A Real Cause — The pattern classifier expects link bursts to correlate with news or viral events. If you run a campaign that legitimately spikes links, pair it with the real-world event so the spike reads as natural rather than coordinated.
- Time On A Good Page Compounds — A page that has existed for years earning links at a steady rate carries an established-authority signal newcomers cannot replicate quickly. Keep authoritative URLs live and consistent rather than churning or re-slugging them.