Most website owners treat their best blog posts as permanent assets. Publish once, rank forever. That model still holds in traditional Google search — a well-optimized page from 2022 can still pull traffic in 2026.
AI search doesn’t work that way.
Perplexity’s algorithm weights content freshness at 40% of its citation ranking signal, according to 2026 citation analysis by Gander and AuthorityTech. Content published within the last 30 days is cited at 3.2x the rate of equivalent pages older than 90 days. The citation half-life — the point at which a page’s citation rate drops by 50% from its peak — is approximately 13 weeks on Perplexity.
That means if you published a solid GEO-optimized post in June, it may already be invisible to Perplexity by September — not because of quality, but because of age.
Here’s exactly what’s happening, why it matters differently across platforms, and the five-part refresh protocol that reverses the decay.
Why Freshness Signals Differ Across AI Search Platforms
The three dominant AI search tools — Perplexity, ChatGPT Search, and Google AI Overviews — use fundamentally different citation architectures. Treating them as one channel is the first mistake.
Perplexity runs real-time web retrieval on every query. It has no pre-crawled index equivalent to Google’s; it fetches, parses, and ranks sources at query time. This makes recency structurally central. A 2026 study tracking Perplexity citations across 14 weeks found that 82% of cited pages were published within the last 30 days. The 40% freshness weighting is the highest of any platform analyzed.
ChatGPT Search operates on a hybrid model: a base of training data (with a knowledge cutoff) combined with Bing-powered web search for live queries. It favors authoritative, established sources — Wikipedia appears in 47.9% of ChatGPT citations, according to Peec AI’s March–June 2026 dataset. Freshness matters but ranks below domain authority and topical depth. A well-cited 18-month-old guide can still perform on ChatGPT while completely vanishing from Perplexity.
Google AI Overviews and the newer AI Mode (launched in full in 2026) each have their own patterns. Despite answering similar queries, AI Overviews and AI Mode cite the same URLs only 13.7% of the time. AI Overviews leans on pages that rank in positions 1–10 (58% citation rate at position 1, dropping to 14% by position 10), while AI Mode pulls from a broader authority pool. Both reward freshness, but less aggressively than Perplexity — BrightEdge data shows only 17% of AI Overview citations now overlap with top organic rankings, down from 76% in mid-2025.
The 13-Week Decay Curve
Gander’s longitudinal analysis tracked 1,400 URLs across Perplexity citations from January to August 2026. The pattern is consistent:
- Week 1–2: Newly published content enters the citation pool. Pages with structured data and clear entity signals index fastest.
- Week 3–8: Citation peak. Pages with 19+ statistical data points receive 2–3x more citations during this window than text-heavy pages.
- Week 9–13: Citation rate begins declining unless the page receives updates, new inbound links, or fresh engagement signals.
- Week 14+: Without intervention, citation rate has dropped ~50% from peak. By week 20, most non-refreshed pages see citation rates below 15% of their original level.
This isn’t a flaw — it’s how Perplexity’s real-time retrieval works. A 2026 analysis from AuthorityTech found that an updated page can move from uncited to actively cited within a single Perplexity index cycle, which runs approximately every 3–5 days for indexed domains.
What “Fresh” Actually Means to Perplexity’s Algorithm
Freshness isn’t simply publishing date. Perplexity’s algorithm evaluates several recency signals, and most can be updated on existing content:
Publication timestamp: The most direct signal. WordPress and most CMS platforms emit this via schema markup. If your post has a datePublished schema property from 2024 and no dateModified, Perplexity treats it as 2024 content regardless of when you last edited the body.
dateModified schema: Updating this property — combined with substantive content changes — is the primary signal that refreshes a page in Perplexity’s citation pool. Surface-level edits (fixing typos) do not reliably trigger re-evaluation. Adding a new data point, a new section, or updating at least 15–20% of the content body does.
Inbound link recency: New backlinks signal to Perplexity that external sources consider this page current. A page with one new inbound link from the past 30 days has a measurably higher citation rate than an identical page with no recent inbound links, per Contently’s 2026 Perplexity optimization report.
Engagement signals via crawl behavior: Pages that receive increased crawl frequency from AI bots (GPTBot, PerplexityBot, Google-Extended) during a 14-day window see higher citation rates. High crawl frequency is itself a signal of content that other sources are citing and linking.
The 5-Part Content Refresh Protocol
Running this protocol on your best-performing GEO content every 8–10 weeks maintains citation rates across the 13-week half-life cycle.
1. Update the data layer. Find every statistic in the post. Search for its 2026 equivalent. Even replacing two or three numbers with fresher equivalents — and noting the date of the source inline (“per BrightEdge Q3 2026”) — signals substantive freshness. Pages with data points that include specific dates in 2026 cite 1.8x more frequently than pages with undated statistics, per Omnibound AI’s citation database.
2. Add a “What’s changed” block. Insert a dated update block at the top of the post: “Updated September 2026: [two sentences on what changed since original publication].” This serves both as a user signal and as a machine-readable freshness marker. Perplexity’s parser treats inline date references as strong recency signals.
3. Update schema markup. Set dateModified to the current date in your Article or BlogPosting schema. In WordPress with Rank Math or Yoast, this updates automatically on republication if the plugin is configured to emit schema. Verify via Google’s Rich Results Test or schema.org validator before republishing.
4. Expand one section with new content. Add a minimum of 200 words to one existing section or create a new section addressing a development that occurred since original publication. A new H2 section titled “What Changed in Q3 2026” or “The September 2026 Update” directly signals recency to both Perplexity and Google AI Overviews.
5. Republish with internal linking. After refreshing, add a link from your most recent (past 30-day) post pointing to the refreshed article. Internal links from new content to refreshed content accelerate re-crawl and reinforce recency signals across your site’s link graph.
Building a Freshness-First Publishing Calendar
The 13-week half-life implies a specific publishing cadence for AI visibility. Rather than one-and-done publishing, the highest-citation sites in Omnibound AI’s 2026 dataset follow a 70/30 model:
- 70% new content: Original posts targeting new angles, current events, and emerging topics. These feed the 30-day citation surge window.
- 30% content refreshes: Scheduled updates to your 10–15 highest-value pages, rotating every 8–10 weeks. Each refresh resets the decay curve.
For most sites publishing 3–5 posts per week, this means 1 refresh post per week alongside new content. The refresh is typically faster to produce than new content (1–2 hours vs. 4–6 hours for a fully original post) because the structure and research base already exist.
Prioritize refreshes based on historical citation volume, not just traffic. A post that was heavily cited 4 months ago and has since decayed is worth more to refresh than a post that never entered the citation pool in the first place.
The Statistical Freshness Advantage
One pattern that holds across all three platforms: statistic-heavy content ages faster but refreshes better.
Pages with 19 or more specific data points earn 2–3x more AI citations during their peak citation window, per the Princeton-led GEO research. But data-heavy pages also decay faster when those statistics become outdated — Perplexity’s algorithm penalizes pages where cited statistics are traceable to sources more than 24 months old.
The implication: if you have high-performing statistical posts from 2024 or early 2025, they are likely decaying on Perplexity right now regardless of their quality. A targeted refresh replacing 2024 stats with 2026 equivalents can restore citations within one index cycle.
65% of all AI bot crawl traffic in 2026 targets content published within the past year, according to Demand Local’s 2026 agency brief. The bots are actively de-prioritizing older content. Your refresh cadence needs to account for this.
Conclusion: Stop Treating AI Search Like Traditional SEO
Traditional SEO rewards permanence. You build a page, earn authority, and collect traffic for years. AI search — especially Perplexity — rewards recency. A 13-week half-life means your content strategy needs a rotation component, not just a production component.
The practical shift: identify your 10–15 highest-value GEO-optimized posts, track their last update dates, and schedule refreshes every 8–10 weeks. Update the data layer, add a dated update block, fix your schema, expand one section, and republish with a new internal link. That’s a 90-minute investment per post that resets a citation curve that took months to build.
Start by finding out where your current content actually sits in AI search. The LLMagnet AI Visibility Scan shows which of your pages are currently being cited by ChatGPT, Perplexity, and Google AI Overviews — and which have already decayed out of the citation pool. Run a free scan before you decide which posts to refresh first.