Google Analytics shows nothing wrong. The page still ranks. Backlinks are intact. Traffic from search is steady. But run the same prompts you ran three months ago through ChatGPT, Perplexity, or Claude, and your brand is gone from the answers. A competitor is there instead.
This is citation decay, and it is the AEO problem most teams find out about six months too late.
What Citation Decay Actually Looks Like
AI citation loss does not announce itself. The first sign most teams notice is that AI-driven referral traffic quietly stops growing, then starts falling. By then the decay has usually been running for weeks. The pattern is consistent: a page cited weekly in March is down to monthly by August and gone from AI answers by year end. The page looks healthy in every traditional metric.
Ahrefs confirmed this clearly. AI Overview citation overlap with Google's top 10 dropped from 76% in mid-2025 to 38% by early 2026. Ranking in traditional search no longer predicts citation in AI answers. The two channels are diverging fast.
Why AI Engines Drop Pages They Used to Cite
There are four main causes.
Recency signals. AI retrieval systems weight content freshness heavily because freshness is a proxy for accuracy. When an engine finds two pages covering the same topic, the one updated more recently tends to win the citation slot. Amsive's 2026 benchmarks show 50% of AI citations going to content updated in the past 13 weeks. AirOps reports a citation penalty of more than 3x for content older than three months without any update. The freshness signal is sent through dateModified in your Article JSON-LD schema, the lastmod field in your XML sitemap, and any visible "last updated" label in the page body.
Stale statistics and outdated claims. A page anchored to a 2023 statistic loses citation share to a page anchored to a 2025 or 2026 equivalent, even when the underlying claim has not changed. AI systems cross-reference claims against multiple sources. When your key statistic has been superseded by newer data published elsewhere, the engine starts preferring the page with the current number.
Vocabulary drift. If your page uses 2022 terminology and the query uses 2026 terminology, the embedding distance between your content and the query can grow large enough that your page never surfaces as a candidate. AI search injects the current year into queries about 28% of the time.
A competitor published something better. When a competitor publishes a thorough, well-structured page on a topic you own, AI engines often rotate within days. Citation share is not a permanent position.
Citation decay accelerates itself. When your brand stops appearing in AI answers, you start losing the community mentions that come from AI-driven discovery. Fewer external mentions reduce AI engines' confidence in your brand as a credible source. Less confidence means further citation loss. Brands that catch a drop in week two spend a few weeks recovering. Those who find it six months later are dealing with a much bigger hole.
How to Fix It: The Content Refresh Sequence
- Update the statistics first. Find every data point on the page. If it cites a study or statistic from more than eighteen months ago, replace it with a current equivalent.
- Refresh the dateModified signal. Update your Article JSON-LD schema so dateModified reflects the actual update date. Update lastmod in your sitemap and republish.
- Add or expand the direct answer block. Every key section should open with a direct, self-contained answer in the first forty to sixty words.
- Audit your vocabulary. Update the language to match how the topic is discussed today.
- Add or expand FAQ schema. FAQPage schema remains one of the highest-impact structured data signals for AI citation.
- Re-promote the page externally. A refreshed page will not automatically recover its citation share. Update internal links pointing to it and share it in relevant communities.
The NotionCue Citation Tracker runs your tracked prompts across all five major engines on a weekly cadence. When your citation rate drops on a tracked prompt, you see it in the dashboard before you see it in traffic.