The 60-Day AI Citation Decay Loop: How Content Freshness Actually Works

The standard advice for AI citation optimization runs something like: update your pages, add the current year to your headings, hit publish. Checked. Done.

That framing treats content freshness as a project. Our research shows it's a cycle -- and the cycle has a specific rhythm that most businesses aren't running on.

The 3.2x citation advantage for content updated within 30 days

In our July 26, 2026 investigation of citation freshness signals (session 88, `research-vault/knowledge/content-freshness-and-citation-signals.md`), three independent data points converged on the same figure.

The ConvertMate 2026 AI Visibility Study, analyzing 80 million citations across platforms, found that content updated within the last 30 days earns 3.2x more AI citations than older content on average. Kevin Indig's 2026 analysis of 1.2 million AI citations arrived at the same figure through a separate methodology. GrowByData's 2026 study matched again.

Three studies. Three methodologies. One number.

The same ConvertMate dataset also produces a higher-magnitude comparison: content updated within 30 days earns 6x more citations than content older than 12 months. Both figures are valid -- the 6x compares the freshest tier to the stalest; the 3.2x compares the freshest tier to all older content combined. The practical point is the same: a page actively updated this month carries a 3.2x citation advantage over a functionally identical page last touched six months ago.

That advantage does not persist.

What happens at 60 days

The Scrunch and Stacker joint study -- 3.5 million citation events across 120,000+ domains and six platforms, run from September 2025 through March 2026 using survival curve analysis with 200 bootstrap resamples -- documented the decay mechanics directly.

Median citation half-life: 4.5 weeks across all platforms. ChatGPT is the most aggressive churner at 3.4 weeks. Perplexity is the most durable at 5.8 weeks.

Our July 26, 2026 session 88 synthesis of multiple 2026 practitioner analyses puts the decay curve in granular terms: citation probability begins declining after 30 days without an update. It drops sharply at 60 days. By 90 days, a page has returned to near-baseline citation probability.

The 60-day drop point corresponds to when the bulk of a citation cohort has churned out -- consistent with the Scrunch/Stacker half-life data. This is not a slow, linear fade. There is a cliff at two months.

For businesses that updated their service pages "a few months ago" and are watching citation rates fall since: that's the cliff.

Why 44% of AI citations are one-time events

Our July 16, 2026 review of the Writesonic 23-million-source study (session 79, `research-vault/knowledge/content-freshness-and-citation-signals.md`) documented a figure that explains why some content falls off the cliff permanently.

Forty-four percent of all cited pages appeared in AI responses exactly once before disappearing from the citation pool entirely. These are transient citations -- picked up once, never retrieved again.

The Writesonic analysis identified four attributes that consistently separate durable citations -- the 56% that persist -- from transient ones.

**Original data.** Pages publishing proprietary statistics or benchmark numbers that exist nowhere else. AI platforms cite their sources. If a page is the only source for a specific figure, the platform must cite it -- there is no alternative.

**Named-expert commentary.** Quotes or assessments attributed to a named, credible person. This gives the retrieval system a human authority to cite alongside the content, increasing trust signals.

**Deep specificity.** Content answering at a use-case level rather than a category level: specific industries served, city names, exact service area, named project outcomes, years operating in a particular market.

**Strong third-party validation.** The page is referenced by authoritative third-party sources -- directories, trade associations, review platforms, media coverage. These third-party references sustain citation pool presence even as the page itself ages between refreshes.

The generic "tips for hiring a plumber" post hits none of these correctly. It is not original data. It has no named authority. It is not specific to a single business. It has no reinforcing third-party presence. It enters the citation pool once, generates a transient citation, and cycles out.

A page with specific project outcomes, an attributed quote from the owner, a service area defined to exact geography, and active directory listings reinforcing the same entity signals is a different object. The decay curve still applies -- but when it falls, it climbs back.

A tiered refresh calendar

Our July 26, 2026 session 88 synthesis describes a content-type approach to freshness maintenance. This is practitioner consensus, not a single primary study -- treat it as directional tactical guidance with a credible mechanism behind it.

**News, trending, and time-sensitive content: 24-48 hour update cycle.** Perplexity re-crawls rapidly and weights recency heavily for time-sensitive queries. Content covering recent developments or seasonal urgency needs updating within 48 hours of material changes to stay in Perplexity's working pool. For standard local service content -- HVAC maintenance descriptions, plumbing service pages -- this cadence does not apply. The short cycle matters for time-sensitive topics only.

**Evergreen competitive content -- service pages, location pages, how-to guides: 60-day loop.** This is the refresh cycle that matters for most local service businesses. The 60-day loop aligns directly with the decay data: content that hasn't been updated in 60 days has hit the sharp-drop point of the decay curve. Refreshing every 60 days keeps these pages in the upper portion of their citation lifecycle consistently.

**Reference and foundational content -- methodology pages, about pages, anchoring resources: quarterly.** These pages build citation authority over time rather than depending on recency. Quarterly timestamp refreshes and content additions -- a new project outcome, an updated client count, a current statistic -- are enough to maintain their position without the more frequent cycle.

Most local service businesses have one content category that determines AI citation performance: their service pages. The 60-day loop is the relevant framework.

What "refresh" actually means for retrieval systems

Our July 26, 2026 investigation also documented a specific failure mode: updating the meta timestamp without refreshing visible content.

Three freshness signals matter to AI retrieval systems: the `article:published_time` and `article:modified_time` meta tags, JSON-LD structured data (datePublished, dateModified), and the visible on-page timestamp. Retrieval systems read all three. A page where the meta tag shows a current date but the visible content still references events or statistics from two years ago creates a mismatch. Some retrieval systems discount or ignore meta-only updates.

The practical checklist for a genuine content refresh: update the visible on-page date or "last updated" notation. Update the meta modified_time tag. Update the JSON-LD dateModified field. Add at least one substantive content change -- a recent project mention, a 2026 statistic, a service area update -- so the freshness signal has something real behind it.

A meta tag update alone is a freshness claim. A meta tag update with changed content is a freshness signal.

The maintenance problem most businesses don't have a system for

A one-time content audit produces a sprint. A 60-day refresh loop requires a calendar.

Most local service businesses don't lack the ability to update their content. They lack a system that tells them when to. Three months pass without a service page update because nobody tracked the last update date, nobody owns the task, and nothing visibly broke.

The decay data says something did break. It's just invisible until the citation rate reports it.

Signal Check at sourcepull.ca shows your current AI citation rate by platform, with a per-platform breakdown of what's being cited and what's not. If your citation rate has dropped since your last check, the 60-day decay clock is the first variable to rule out before diagnosing anything more complicated.