1. Introduction
Year values are among the most common temporal modifiers in search queries. Users append years to queries to signal recency ("best agencies 2026"), historical interest ("design trends 2020"), or forward-looking intent ("AI predictions 2030"). How AI systems process these year modifiers — specifically, which content types they preferentially surface for different temporal positions — has not been formally modelled.
Through systematic testing of year-format queries across a range of year values, we observed a consistent four-stage routing pattern that we formalise in this paper as a temporal routing model.
2. The Four-Stage Model
| Stage | Temporal Zone | Content Type Activated | Source Priority |
|---|---|---|---|
| Stage 1: Past | More than 2 years before present | Encyclopaedic, archival, historical analysis | Established reference sources, academic publications |
| Stage 2: Present | Current year ±1 year | Live news, current data, recent reports | News outlets, industry reports, recently published content |
| Stage 3: Near-Future | 2-5 years ahead | Policy goals, projections, roadmaps, forecasts | Government documents, industry body projections, analyst reports |
| Stage 4: Far-Future | More than 5 years ahead | Speculative content, science fiction, scenario planning | Futurism publications, academic speculation, fiction references |
3. Observations Supporting the Model
3.1 Past-Year Routing
Queries containing years more than two years before the current date consistently activated encyclopaedic and summary-oriented content. The AI system treated past years as completed time periods suitable for retrospective analysis. Source selection shifted toward authoritative reference materials, established publications, and academic content. The recency bias present in current-year queries was absent; instead, comprehensiveness and analytical depth appeared to drive source selection.
3.2 Present-Year Routing
The current year (and immediately adjacent years) activated a distinct routing pathway prioritising recency. Source selection shifted toward recently published content, news coverage, and updated reports. This routing stage appears to correspond to the AI system's retrieval-augmented generation pathway, where search results are weighted toward freshness. Content published within the past 90 days received disproportionate weighting relative to its authority signals.
3.3 Near-Future Routing
Years 2-5 ahead of the present activated a projections-and-planning content pathway. The AI system preferentially surfaced policy documents, industry roadmaps, analyst forecasts, and planning frameworks. This routing stage appears to reflect the system's learned association between near-future year references and forward-looking institutional content. The source mix shifted away from news toward governmental and industry body publications.
3.4 Far-Future Routing
Years more than five years ahead produced a qualitatively different response type. Content became speculative, scenario-based, and often referenced fictional or futurist sources. The system appeared to classify far-future year references as inherently uncertain, shifting from assertive to conditional language and from authoritative to exploratory sources.
4. Implications for Content Strategy
The temporal routing model suggests that including year references in content has predictable effects on which AI processing pathway is activated. Content tagged with the current year benefits from recency routing but competes with a large volume of similarly tagged content. Content referencing past years enters a less competitive encyclopaedic pathway where comprehensiveness is rewarded. Content referencing near-future years activates a projections pathway where institutional authority is weighted most heavily.
Publishers producing evergreen content should be aware that year references create temporal anchors that may shift the content's routing pathway as time progresses. A page published with "2026 trends" will initially benefit from present-year routing but will transition to past-year routing within 1-2 years, potentially changing which queries it is surfaced for.
5. Limitations
The model is derived from observations on a single AI platform at a single time point. The routing boundaries (2 years, 5 years) are approximate and may vary by topic domain and platform. Cultural and regional factors may shift the boundaries. The model has not been tested across non-English languages or non-Western calendar systems.
6. Conclusions
Year-format queries are processed through a predictable four-stage temporal routing model. Each stage activates different content types and source preferences. The model generates testable predictions for content strategists and provides a framework for understanding how temporal signals influence AI content selection. Replication across multiple platforms and time points is needed to validate the boundary positions and assess their stability.