Half of all AI-cited content is under 13 weeks old. That figure appears consistently across independent studies — Amsive's analysis of 50,000 URLs, AuthorityTech's 2026 freshness report, and Radiant Elephant's synthesis of 17 million citations all converge on the same 13-week window as the peak AI citation eligibility zone.
A developer on DEV Community logged their blog's AI citations every week for nine weeks. The citations peaked at week three. By week nine they had fallen by more than half. Their Google organic traffic for the same pages never moved. The two signals run on different clocks — and the AI citation clock is much faster.
What makes this uncomfortable is that the decay is not a quality problem. The pages did not get worse. The pool they compete in got younger. AI retrieval systems bias toward freshness as a proxy for accuracy — a page with a recent dateModified is statistically less likely to contain stale facts than a page last touched six months ago. That heuristic produces the decay pattern whether your content is genuinely outdated or simply unedited.
The systematic fix is a freshness maintenance calendar, not a panic refresh cycle. Brands that maintain citation rates over time treat content like infrastructure: they schedule updates, rotate through high-value pages on a cadence matched to each platform's freshness half-life, and measure whether updates are resetting the citation clock. This post covers exactly how to build that system.
What Is the Citation Half-Life for Each AI Engine?
Freshness preference is not uniform across AI engines, and a single maintenance cadence serves none of them well. The platform-specific data from ICODA's April 2026 citation decay analysis sets the baseline.
ChatGPT has the shortest citation half-life: approximately 3.4 weeks. This is the most aggressive recency bias in the category. 76.4% of ChatGPT's most-cited pages were updated within the last 30 days. ChatGPT is also the dominant source of AI referral traffic — accounting for 87.4% of all AI-referred sessions in Q1 2026 per Conductor. The combination of extreme recency bias and high traffic volume means ChatGPT-facing content needs the most frequent updates. Brands that only refresh quarterly are leaving ChatGPT citations on the table every month they wait.
Perplexity has the longest citation durability: approximately 5.7 weeks. Perplexity does real-time web retrieval on every query, which means it has access to fresh content immediately but can also surface older content when it scores highly on relevance and structure. Perplexity is 68% more forgiving than ChatGPT on freshness. However, approximately 50% of Perplexity citations still come from current-year content. Six-week refresh cycles on priority Perplexity-facing content are the practical minimum.
Google AI Overviews show the weakest freshness bias. Citation patterns on Google AI Overviews track more closely with organic ranking age profiles than the other engines do. Pages that are already in the top 10 for a query are more likely to be cited in AI Overviews regardless of recency. However, the gap narrows on time-sensitive queries — statistics, market data, and tool comparisons decay faster even in Google AI Overviews as newer sources publish.
Claude and Gemini sit between Perplexity and Google AI Overviews. Both show measurable freshness preference without the extreme bias of ChatGPT. For most brands, a 6 to 8 week refresh cadence covers Claude and Gemini adequately when ChatGPT and Perplexity have faster cycles in place.
What Counts as a Substantive Update for AI Freshness Signal?
This is where most content teams make the mistake that kills their freshness strategy. Changing a publish date without changing content does not reset the freshness signal. AI systems compare current page content to cached versions and can detect cosmetic updates. Google's John Mueller explicitly confirmed that date changes without substantive content changes provide no freshness benefit. The same principle applies to AI retrieval systems.
Three types of updates that AI systems treat as genuine freshness signals:
Statistical refresh. Replace any statistic more than six months old with the most current available figure. Name the new source and date inline: "Per Ahrefs analysis of 17 million citations (July 2026)" rather than "per recent research." The Princeton GEO study found adding named, dated statistics is the highest-leverage single AEO tactic — combining it with a dateModified update delivers both the content improvement signal and the freshness signal simultaneously.
Section addition addressing a new subtopic. Add a new H2 section covering a question that has emerged since the original publication date. This adds word count, adds a new heading entity for AI engines to match against queries, and creates a genuine content difference that crawlers comparing current and cached versions detect as a substantive change. The new section should be at least 150 words and should address a query the original article did not explicitly cover.
Schema dateModified update paired with IndexNow ping. Every substantive content update should be followed immediately by a dateModified schema update and an IndexNow ping. DateModified tells AI engines when the content last changed at the machine-readable level — without it, the content improvement is visible to humans but invisible to schema parsers. IndexNow notifies Bingbot (which feeds ChatGPT through Bing) and other supporting engines within minutes of the update. The full IndexNow configuration process is in the Copilot and Bing AEO guide.
How Do You Build a Freshness Maintenance Calendar Without Overloading Your Team?
The mistake teams make when they first understand citation decay is applying equal maintenance to all content. Ninety pages on a quarterly refresh schedule is a 30-page-per-month workload before a word of new content is written. That is unsustainable and unnecessary.
A tiered refresh schedule based on business value and decay risk concentrates effort where it produces the most citation return.
Tier 1 — High-value, fast-decay pages. Your pillar pages, comparison guides, and statistics-heavy posts. These are the pages that earn the most citations and decay fastest because they cite time-sensitive data. Refresh cycle: every 4 to 6 weeks. The update does not need to be a full rewrite — one new statistic, one updated data point, or one new FAQ question is enough to reset the freshness signal. Assign a rotating ownership across your content team so no one person owns the full Tier 1 load.
Tier 2 — Medium-value, medium-decay pages. Tutorial posts, how-to guides, and implementation articles. These decay more slowly because the information is more durable. Refresh cycle: every 8 to 10 weeks. Focus updates on examples that may have become outdated, tool versions that have changed, and FAQ sections where new questions have emerged from support tickets or community discussions.
Tier 3 — Lower-value or highly durable evergreen pages. Foundational definitions, historical case studies, conceptual explanations. These decay slowest and have the lowest citation volume to protect. Refresh cycle: quarterly, or triggered by a competitor publishing newer content on the same topic. A competitive trigger from your weekly prompt tracking is the most efficient signal — if a newer page starts beating you on a Tier 3 prompt, escalate that page to Tier 2 and refresh it.
The AEO content team workflow guide covers how to integrate the freshness maintenance calendar into a broader weekly and monthly content rhythm. The key operational principle: maintenance tasks should appear on the same calendar as production tasks, not on a separate "someday" list. Pages you are not actively maintaining are losing citations to competitors who are.
How Do You Prioritise Which Pages to Refresh First?
Page prioritisation uses four criteria that predict the highest citation return per hour of maintenance effort.
Current citation rate. Pages that already earn citations are the highest priority for freshness maintenance — they have proven citation eligibility and are most at risk of losing it as they age. Pages with zero citation rate are not freshness problems; they are structural or content problems that freshness maintenance will not fix. A page earning 40% citation rate on its target prompt five weeks ago that has now dropped to 15% is a classic decay signature — refresh it.
Age of statistics in the content. Pages whose most recent statistic is older than six months are at higher risk of freshness decay than pages with recent data. Run a quick manual audit: open your top 20 pages and check the most recent date attributed to a statistic. Pages where all statistics are older than Q4 2025 are the first refresh targets.
Competitor freshness advantage. If your weekly prompt tracking shows a competitor page was published or updated in the last 30 days and is now citing above you, that competitor has a freshness advantage on that prompt. Refreshing your page to add a newer statistic and a new section can close the gap within two to three weeks on Perplexity and four to six weeks on ChatGPT.
Revenue or pipeline value. Pages that generate direct leads or product trials through AI referral traffic are the highest business-value refresh targets. Check GA4 for AI referral sessions by landing page. The pages generating the most AI-referred conversions earn the most maintenance priority, regardless of their current citation rate — losing those citations means losing that pipeline.
How NotioncCue Helps You Detect and Prevent Citation Decay
Citation decay is invisible in GA4 until it produces a traffic drop — and by that point, a competitor has been in the citation position for weeks. The decay starts showing in prompt tracking data before it appears in any traffic metric.
The NotioncCue Prompt Tracker runs your target prompts weekly across all five engines and records citation presence, competing sources, and the specific URL cited each week. Week-over-week comparison shows you exactly when a page drops from a cited position — the week the decay begins, not the week traffic falls as a consequence. A page that earned a citation on your target prompt for four consecutive weeks and then dropped out in week five is a freshness decay signal. It tells you the page needs a substantive update before a competitor locks in the position.
The NotioncCue Citation Tracker monitors what AI engines say about your brand on a weekly cadence across all five engines. It catches a second freshness risk that prompt tracking alone misses: stale brand descriptions. When AI engines are citing your brand in answers but describing you with information from six months ago — an old product name, a discontinued feature, a pricing model that has changed — that is a training data and retrieval freshness gap. The Citation Tracker surfaces these discrepancies so you can correct them at the source (schema, documentation, review platform profiles) before they persist in AI-generated buyer answers for another model update cycle.
Start your free NotioncCue trial and set up weekly prompt tracking across your top 15 target prompts. You will see your citation decay pattern within four weeks — which pages are holding, which are declining, and which competitors are taking your positions while you are not watching.
The most common freshness mistake is updating a dateModified timestamp without submitting for re-crawl. A dateModified update that is not followed by an IndexNow ping or a Google Search Console URL inspection request can take weeks to be picked up by AI crawlers on their standard crawl schedule. Submit every updated page for re-crawl immediately after the update — this is the difference between a freshness signal that lands within 48 hours and one that lands three weeks later, by which time the competitive window may have closed.
Frequently Asked Questions About AEO Content Decay and Citation Freshness
Does content decay affect all AI engines at the same speed?
No. ChatGPT decays fastest, with a citation half-life of approximately 3.4 weeks. Perplexity is most forgiving at approximately 5.7 weeks. Google AI Overviews are least freshness-sensitive for established pages already ranking in the top 10. If you can only maintain one engine's freshness requirements, prioritise ChatGPT-facing content because ChatGPT generates the most AI referral traffic — 87.4% of all AI-referred sessions — and has the most aggressive recency bias.
Is there a way to prevent decay on pages you cannot update frequently?
Two structural approaches slow decay without requiring frequent content updates. First, build your content around durable frameworks and methodologies rather than specific statistics that change frequently. A page explaining the principles behind AEO citation selection decays slower than a page listing specific citation rates that change monthly. Second, ensure your page has strong third-party validation signals — backlinks, Reddit mentions, and review platform references — that persist independently of your own freshness maintenance. Authority signals from third-party sources counteract freshness decay in AI engines that weight both freshness and authority, preventing complete displacement even when your content is 90 days old.
How do you know if a content refresh has actually reset the citation clock?
Run your target prompt on Perplexity three to five days after the refresh and compare the citation result to the pre-refresh baseline. Perplexity responds fastest to freshness updates — often within 48 to 72 hours of a page being re-crawled via IndexNow. If the citation has returned or improved on Perplexity within that window, the refresh has been detected and the freshness signal has been reset. For ChatGPT, the same comparison should be run at two and four weeks post-refresh, as ChatGPT's retrieval cycle is slower. Track both in the NotioncCue Prompt Tracker to get a before-and-after citation rate comparison without manual logging.