How should teams interpret short-lived citation spikes?
Mostly with caution. A spike can be real and still fail to persist because 73.5% of citations are one-and-done and the median URL lifespan is 0 days.
The key numbers.
- One-and-done citations
- Citations that appear once and vanish.
- 73.5%
- Median citation lifespan
- Most citations disappear immediately.
- 0 days
- Still active citations
- Small share of URLs still active at the end of the window.
- 6.8%
The right question is not whether a spike occurred, but whether the visibility survived repeated observation over time.
The evidence.
| Measure | Result |
|---|---|
| One-and-done citations Citations that appear once and vanish. | 73.5% |
| Median citation lifespan Most citations disappear immediately. | 0 days |
| Still active citations Small share of URLs still active at the end of the window. | 6.8% |
Population and method
Built from 857,138 reports, 108,650 citations, 10,991 brands, and 8 tracked models across a 10-month observation window.
Where the evidence stops
Decay describes observed persistence, not the full causal mechanism behind why a citation disappears.
Brand- and domain-level averages can hide very different retention patterns by query class or model.
Some of the stickiest or most volatile domains come from niche query pockets and should be read as examples, not universal leaders.
What to do with this
- Measure visibility as a moving system rather than a one-time citation snapshot.
- Refresh and monitor citation-driving pages on the cadence your models actually decay.
- Separate durable wins from temporary spikes to prevent overreacting to short-lived mentions.
This turns study findings into an operating rule teams can use to decide what to publish, refresh, or measure next, avoiding overinvestment in transient visibility.
Questions about this answer.
Cite this answer.
Source study: Citation Decay“Mostly with caution. A spike can be real and still fail to persist because 73.5% of citations are one-and-done and the median URL lifespan is 0 days.”
Mack Grenfell · How should teams interpret short-lived citation spikes?
The Half-Life of AI Citations · Trakkr Research · Updated 30 March 2026