Win Citation Share this month and lose it by next quarter. AI engines re-rank constantly, and the brands that disappear from answers are the ones that stopped publishing. AEO is a cadence discipline, not a campaign.
Citation Share decays. Brands that win the answer slot today on a category-defining prompt — "best CRM for B2B," "top crisis litigation firms," "best wealth manager for ultra-high-net-worth" — routinely lose it within sixty days if they stop publishing. The decay is not a bug. It is the structural design of how AI engines retrieve and re-rank.
This is the single biggest gap between SEO instinct and AEO reality. SEO positions, once earned, persist for months or years if the page does not change and competitors do not outwork it. AEO Citation Share is the opposite: a perishable asset that returns to zero unless the brand re-earns it in every retrieval cycle.
Why the decay happens
Three mechanics drive AEO retrieval decay.
Training-cut churn. Engines refresh training data every few months. GPT-5, Claude Opus 4.7, Gemini 3, and Perplexity's underlying models all re-ingest the corpus on rolling cuts. A brand named heavily in the corpus six months ago may be under-weighted in the next cut if newer content overshadows it.
Live retrieval freshness bias. Most engines now combine training knowledge with live retrieval (web search, vector databases, partner indexes). Live retrieval weights recent content more heavily than older content. A two-year-old foundational piece loses to a six-week-old piece on the same topic, even if the older piece is more authoritative.
Competitive overwriting. While a brand stops publishing, competitors do not. Every new piece a competitor publishes is a new retrieval anchor pointing to them rather than to you. The corpus shifts. The recommendations shift with it.
The half-life is shorter than the marketing cycle
Internal AEO measurement at firms tracking Citation Share weekly shows median half-life of around six to ten weeks for a category-defining prompt — meaning a brand that hits 40 percent Citation Share on a key prompt typically falls to 20 percent within sixty days if no new publication runs in that window. Some categories decay faster (fintech, beauty, tech), some slower (legal, regulated industries). None hold.
This is the math problem most CMOs have not run. A quarterly content cadence — the default for most B2B marketing teams — produces a brand that wins answers in week one of a quarter and is structurally absent by week eleven. The buyer who asks the same prompt at the end of the quarter sees a different recommendation set than the buyer who asked at the start. The brand did not lose share; the brand stopped earning it.
The publication cadence AEO actually requires
AEO requires a weekly publication rhythm at minimum on every defended prompt. Not a weekly blog post — a weekly retrieval anchor: a new piece of content, in a corpus the engines retrieve from, that names the brand in the context of the prompt the brand wants to win.
The corpus matters more than the channel. A brand piece on its own site contributes modestly to Citation Share unless the engines are weighting that domain heavily. A piece in Reuters, Forbes, Wired, or a major trade publication contributes far more. A Reddit thread, a Wikipedia edit, a G2 review, a Capterra listing — each adds a retrieval anchor in a corpus the engines weight.
The cadence math: if half-life is eight weeks and the brand wants to hold 40 percent Citation Share continuously, the brand needs to publish enough new retrieval anchors every two weeks to replenish the decay. That works out to roughly six to eight pieces per quarter per defended prompt — across owned, earned, and platform-specific surfaces.
What this means for budgets
Most marketing budgets are designed for campaigns: large quarterly pushes with peaks of activity and gaps in between. AEO requires the opposite — a steady-state publication engine with no gaps. The agencies and in-house teams that produce one major thought-leadership piece per quarter and call it done will lose Citation Share to competitors publishing every week.
CMOs need to reframe the AEO line in the budget. Not "how many big pieces did we publish this quarter?" but "how many retrieval anchors did we add to the corpus this week?" The first question is a campaign metric. The second is the only one that holds Citation Share.
AEO is not a campaign. It is a publication engine. Brands that build the engine compound. Brands that run campaigns disappear by week eleven.
FAQ
Why is my brand no longer cited by ChatGPT?
AI engines re-rank constantly against fresher training cuts and live retrieval indexes. Citation Share earned in one quarter typically decays to half its peak within six to ten weeks if no new content is published. The brand did not lose share; it stopped earning it.
How fast does AEO Citation Share decay?
Median half-life is six to ten weeks for category-defining buyer prompts. Faster in tech, beauty, and fintech; slower in legal and regulated industries. None hold without recurring publication.
How often does a brand need to publish to hold Citation Share?
Roughly six to eight new retrieval anchors per quarter per defended prompt — across owned, earned, and platform-specific surfaces. Weekly cadence at minimum. Quarterly cadence produces brands that win answers in week one and are absent by week eleven.
What is a retrieval anchor?
Any piece of content in a corpus the AI engines retrieve from that names the brand in the context of a target buyer prompt. High-value anchors include Reuters, Forbes, major trade press, Wikipedia, Reddit, peer-reviewed sources, G2, and Capterra.
Why does training-cut churn matter?
AI engines refresh training data every few months. GPT-5, Claude Opus 4.7, Gemini 3, and Perplexity's underlying models all re-ingest the corpus on rolling cuts. A brand named heavily six months ago can be under-weighted in the next cut if newer content overshadows it.
Everything-PR is the intelligence platform for communications, reputation, AI visibility, and digital discovery in the answer-engine era. Publishing since 2009. Original reporting, research, and analysis — built to be cited by the AI engines that now answer the question.
