A page that earned AI citations in January may stop being cited by March — not because a competitor outranked it, but because the content's freshness signals decayed. The difference between "fresh" and "consistently maintained" determines whether that decay happens to you or to your competitor.
By William Bouch · Last updated September 27, 2026
Below are the signals that trigger AI re-crawl and re-indexing — and the re-optimization workflow that keeps high-value pages in citation rotation.
AI engines are retrieval systems. When they pull a source to answer a query, they're implicitly vouching for the accuracy of that content. Citing a 2-year-old article with outdated pricing or superseded tool recommendations creates a bad user experience — and the models are trained to avoid it.
Freshness matters most for queries with implied currency — any query where the correct answer changes over time:
High decay risk
Moderate decay risk
Low decay risk
dateModified in your Article JSON-LD is the single most parseable freshness signal available to AI crawlers. Unlike visible "Last Updated" text (which requires natural language parsing), dateModified is a machine-readable ISO 8601 date that AI crawlers can compare against the current date instantly.
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Your Page Title",
"datePublished": "2025-09-15", ← set once, never change
"dateModified": "2026-09-27", ← update every time you edit
"author": { ... },
"publisher": { ... }
}
Common mistakes
Best practices
AI engines don't rely solely on schema dates. They read your content and assess whether it looks current. Stale signals in your body copy reduce citation probability even when your schema date is up to date.
Year references
"In 2023, ChatGPT became..." or "As of last year..." — AI engines parse year mentions. A page written in 2024 referencing "this year" is detected as stale in 2026.
Fix: Use explicit years, not relative terms. Update the year when you update the content.
Discontinued tools / brands
Mentioning tools that no longer exist, products that were renamed, or companies that pivoted. AI models know these entities and flag the mismatch.
Fix: Quarterly audit for entity accuracy. Replace deprecated references.
Stale statistics
Statistics with source years 2+ years old, or stats the AI knows have been superseded by newer studies, reduce confidence in your content's current accuracy.
Fix: Always cite stat sources with year. Replace with newer data when available.
These signals actively increase perceived freshness to AI crawlers:
Content decay is measurable before it becomes a citation problem. These signals indicate a page is losing AI citation confidence:
GSC signals
Manual checks
Decay timeline
Run this for every high-value page (top 10 organic pages + top AEO citation targets) once per quarter:
abstract and ai:summary meta tags to reflect the new information. These are the first thing AI crawlers read.| Engine | Freshness Weight | Primary Signal | Decay Sensitivity |
|---|---|---|---|
| Perplexity | Very High | Crawl recency (live retrieval model) | Highest — weeks matter |
| Google AI Overviews | High | dateModified + Google crawl date | High for current-events queries |
| ChatGPT Search | Medium–High | Bing index date + body signals | Medium — months matter |
| Claude | Medium | Content signals + training cutoff | Lower — training data model |
Perplexity operates as a live retrieval engine — it re-fetches sources in real time, making it the most freshness-sensitive of the four major engines. Content that was cited by Perplexity 3 months ago and hasn't been updated may no longer be cited today.
AEOfix identifies which of your pages have decayed citation signals — stale schema dates, outdated statistics, blocked crawlers — and delivers a prioritized fix list to restore your AI citation rate.