What is citation decay?
Citation decay is a decline in a defined page’s citation measure over comparable observations; a half-life requires an explicit baseline, threshold and time rule.
An old page that is cited today has not necessarily decayed. A previously cited page missing from a later answer may reflect ordinary answer variation. The term becomes useful when it names a measured change through time under a stated observation design.
The practical reporting question belongs in the library’s historical data and time-series guidance. Here we define the event that a time series would need to establish.
Where does the current AI-search usage come from?
The earliest dated primary product explanation of the exact term we verified is Profound’s 29 September 2026 announcement by Joey Adelman and Matthew Huo. That is a bounded source finding, not a claim that no earlier author used the words in another field or on another site.
Profound defines a page’s citation share against all citations across the models it tracks, smoothed over a two-week window. Its half-life runs from the page’s peak to half that peak, with a requirement to remain at or below the threshold for 14 consecutive days. These are Profound’s operational choices, not a universal definition imposed on other studies. [Adelman & Huo, 2026].
The product also excludes pages that do not meet its age, citation-count and peak-activity rules. Consequently, a summary for eligible pages should not be generalized to every published URL. The vendor’s concept is useful precisely because it says which curve and which threshold it means.
Choose the quantity before fitting a curve
Absolute citation count, share of citations, proportion of answers citing a URL and proportion of questions ever citing it are different measures. A page can keep receiving the same number of citations while losing share because the rest of the observed citation pool grows.
Specify the URL identity rule. A changed tracking parameter should not necessarily start a new page history, while a substantially revised document at a stable address may need a version label. Explain whether redirects, canonical declarations or a manually maintained document identifier connect observations.
Define the observation window and treatment of missing collection. An unsuccessful capture is not an observed absence. A page that has not crossed the threshold by the last observation has an incomplete time-to-event record, not a measured infinite life or an invented future decay date.
Why a missing citation is weak evidence by itself
Generated answers can differ under repeated questions. A before/after disappearance might therefore be a transient answer change rather than a durable loss. Repetitions help reveal that variation, while a stable comparison set helps separate page-specific changes from broader engine or collection changes.
Do not select a peak retrospectively and then describe every ordinary fluctuation below it as deterioration. State whether the peak is defined across the whole observed series, within a moving window or from an initial baseline. Different definitions can produce different half-lives from the same records.
Calendar age also differs from time since first observed citation. A document may have existed for years before entering the tracked question set. Without earlier observations, “first cited” means first seen by that collection, not necessarily the first citation anywhere.
Can decay justify a content refresh?
A declining series can identify a page worth investigating. It does not establish that rewriting it will recover citations. Check whether the facts became stale, the question set changed, a competing source appeared, or the engine changed its source presentation.
If an intervention follows, record its date and preserve the earlier version. Report the outcome against an appropriate comparison design. A refresh recommendation and a demonstrated refresh effect are separate claims, and a decay estimate supplies only the former unless the later test supports more.
Joey Adelman and Matthew Huo write, “We track this on a rolling two-week average”. That smoothing choice is part of Profound’s published product definition, rather than a rule established for every citation dataset. [Adelman and Huo, 29 September 2026, “What Citation Decay shows”; verified 30 October 2026].
You can distinguish an ageing document from a measured decline in citation share.
What our own data shows
We have no measurement on this point: citation decay. Study 3 is a registered protocol for release-to-release absence persistence and remains unrun until its comparability conditions are met. Vendor absence persistence is also a different event from a particular URL’s citation half-life.
The September 2026 release supplies the answers for our repeated-question analysis. It observed 1,259 mixed-naming groups among 14,192 complete vendor-question-engine groups: 8.9%, Wilson 95% interval 8.4% to 9.4%. Mixed means named in one or two of three valid repeats. These within-release groups do not measure elapsed-time decay. See the measurement-noise finding and the Study 3 protocol.
Sources
- Adelman, J., & Huo, M. (2026, September 29). Introducing Citation Decay in Profound. Profound. Citation-share, smoothing, half-life and eligibility definitions. https://www.tryprofound.com/blog/citation-decay Verified 30 October 2026.
- Broadcastwell. (2026). Does absence persist from one release to the next? [Registration]. OSF. Registered 30 October 2026, 01:12:44 UTC. https://doi.org/10.17605/OSF.IO/MKJ9V Verified 30 October 2026.
- Broadcastwell. (2026). Absence Index, release 2026-09 [Data set]. Zenodo. Archived answers and release metadata. https://doi.org/10.5281/zenodo.22907695 ; derived repeat-disagreement table. Verified 30 October 2026.
Cite this page
Broadcastwell research team. (2026, October 30). What is citation decay?. Broadcastwell. https://docs.broadcastwell.com/research/glossary/citation-decay/ Published 30 October 2026; updated 30 October 2026.
@online{broadcastwell2026research36,
author = {{Broadcastwell research team}},
title = {What is citation decay?},
year = {2026},
date = {2026-10-30},
url = {https://docs.broadcastwell.com/research/glossary/citation-decay/},
urldate = {2026-10-30},
note = {Published 2026-10-30; updated 2026-10-30}
}
Prepared by the Broadcastwell research team. Lead: Lauren Chen, AI architect and specialist.
Sources checked 30 October 2026. Independent check: 30 October 2026, Broadcastwell verification desk. This page has not been peer reviewed.
Last updated: 30 October 2026. Next review: 30 January 2027.