Citation decay is the gradual loss of AI citations a page or brand was previously earning, typically driven by content staleness, competitor improvements, and AI-platform retrieval-index updates. LLMrefs data shows pages not updated in 90+ days lose AI citations at roughly 3x the rate of regularly-maintained pages. Inversely paired with citation velocity: velocity is gain, decay is loss; the net of the two defines the trajectory.
What is citation decay?
Citation decay is the rate-of-loss side of AI visibility. A page that was cited 50 times a month earlier this year might now be cited only 20 times because AI platforms have moved to fresher or more authoritative alternatives. The page's content itself has not changed; the context around it has. Competitors published better coverage; freshness signals lapsed; the retrieval index re-weighted. Decay is what happens when an AI-visibility asset loses value without active maintenance.
The concept is an adaptation of "content decay" from classic SEO, which itself adapted from "link decay" in link-building. The underlying idea is consistent across all three: a ranking asset (a page, a backlink, an AI citation) is not static. It has a half-life. Without active investment, it depreciates. Citation decay names the AI-search-era version of this dynamic and gives marketers a metric to track.
The practical point: citation count is a stock, not a flow. A brand with 100 monthly citations today is not guaranteed 100 next month. Keeping that 100 requires either sustained citation velocity (adding new citations) or low decay (keeping existing ones). Most programs track velocity and miss decay, which masks the real trajectory.
What causes citation decay
Four drivers, ordered from most-controllable to least.
Content staleness (controllable)
The most common cause. A page's visible update date hasn't moved in a year, the schema dateModified matches the original publish date, and the sitemap lastmod hasn't shifted. AI platforms read all three as signals that the content is aging out. Content freshness is the direct lever; regular refreshes counteract decay almost fully on salvageable pages.
Competitor content improvements (controllable-by-proxy)
A competitor publishes a better version of the content AI was citing from you. Their version is more structured, more recent, more thoroughly cited. AI platforms swap. You cannot directly stop competitors from improving their content, but you can respond by improving yours. This is the kind of decay that rewards active awareness rather than fire-and-forget publishing.
Retrieval-index updates (external)
AI platforms periodically re-crawl their source sets and re-rank what they cite. These updates are opaque from the outside; you see the effect (a drop in citations) without seeing the cause. Most are small and transient; occasional larger index updates cause category-wide shifts. Watch for patterns in citation decay that correlate with known AI-platform version releases.
Training-data cutoff advancement (external)
When an AI platform trains on a newer dataset, older coverage of the same topic gets partially superseded. A brand's 2023 coverage that got cited heavily in 2024 may be cited less in 2026 simply because more recent coverage exists in the new training set. This is structural decay; the only counter is to have sustained coverage across years so the brand's presence survives each training cycle.
How to measure citation decay
Decay measurement requires page-level granularity, not site-wide aggregates.
Track citations per-page over time
AI prompt-monitoring tools that break citation counts down by source URL give you the raw data. For each cited page, track the weekly citation count. A declining trend is the per-page decay signal.
Compute decay rate
For a given page, decay rate is the percentage drop in citation count per unit of time. A page that went from 20 citations/month to 10 over 3 months has a -50% quarterly decay. Benchmark against your category's typical decay curve.
Segment decay by driver
If decay correlates with content freshness lapses (dateModified hasn't moved in 6+ months), the driver is staleness. If decay correlates with a competitor publishing a new page on the same topic, the driver is competitor improvement. Different drivers require different responses; mixed drivers require both.
How to counteract citation decay
Three moves, ordered by leverage.
Schedule content refreshes
The single highest-impact counter. Refresh high-value pages every 60-180 days depending on topic velocity. Update the visible date, the schema dateModified, the sitemap lastmod, and the substantive content. Skipping the substantive update makes the refresh a cosmetic signal that AI platforms increasingly detect and discount. Covered in depth in the content freshness entry.
Upgrade high-potential pages
Pages that earn citations but could earn more usually have structural gaps: missing schema markup, weak direct-answer paragraphs, stale internal links. Upgrading these signals on a page converts passive decay into active re-earning.
Retire and consolidate
Not every page is worth fighting to save. Pages with weak topical depth, low traffic, and declining citations often decay because the underlying content was never strong. Consolidating several weak pages into one strong topic cluster pillar produces more durable citation signal than trying to resurrect each page individually.
Measurement closes the loop. Our AI Automation module tracks per-page citation counts over time and flags decay patterns for marketer review (one operational use of Marketer in the Loop).
Common misconceptions
Citation decay only happens to old content
Older content decays faster, but newer content decays too. A page published three months ago may already be losing citations if a competitor just published something better. Decay is a function of relative value against the current retrieval set, not purely of absolute age.
A high-velocity program doesn't need to worry about decay
New citation gains can mask decay happening on older pages. A brand that's adding 10 citations a week while losing 8 has net +2 velocity; without tracking decay you'd think +10. When decay sources become known (staleness, competitor moves), the same program can focus on preventing losses rather than only chasing new gains.
Refreshing the date stamp is enough
AI platforms increasingly detect "zombie refreshes" where only the date moves without substantive content changes. The signal gets discounted, and the decay continues. Real refreshes require real edits.
Frequently asked questions
#What is citation decay in simple terms?
Citation decay is the gradual loss of AI citations a page or brand was previously earning. A blog post that was cited 50 times a month in January might be cited only 20 times in April because AI platforms have moved to fresher or more authoritative alternatives. Decay is the AI-search-era equivalent of classic content decay: a ranking asset loses value over time unless actively maintained.
#How fast do AI citations decay?
LLMrefs data shows pages not updated in 90+ days lose AI citations at roughly 3x the rate of regularly-maintained pages. The exact curve is category-dependent - news and tech content decay faster than evergreen reference content - but 90 days is a reliable inflection point. Past that threshold, decay accelerates unless the page is refreshed or replaced.
#What causes citation decay?
Four drivers: (1) content staleness (the page's dateModified signal is old, freshness signals drop), (2) competitor content improves and AI platforms swap citations, (3) AI platforms re-index their retrieval caches with new crawls, (4) training-data cutoffs advance and the brand's older coverage gets superseded by newer sources. The first two are under the brand's control; the second two are external.
#How do I know if my content is decaying?
Track citation velocity alongside share of AI voice. A brand with flat share of AI voice but declining per-page citation counts is experiencing decay that's being masked by new citations elsewhere. Prompt-monitoring tools should break citations down by source page so you can see which specific pages are losing ground. Page-level decay data is the diagnostic; site-wide averages hide it.
#How do I counteract citation decay?
Three moves in order of leverage. First, schedule content refreshes on a 60-180 day cadence depending on topic velocity. Second, upgrade high-value pages with current schema, direct-answer paragraphs, and freshness markers. Third, let pages that can't be salvaged retire - consolidate into stronger pillar pages rather than letting stale URLs drag the domain down. Fighting decay on every page is infeasible; fighting it on the 20% that matter most is standard practice.
