Content Freshness in AI Search: Why Older Pages Respond Better

Content freshness is good AEO advice. A July 2026 Seer Interactive primary study of 7,683 pages and 47,097 citations confirms it: freshly updated pages earn more AI citations than stale ones, across ChatGPT, Perplexity, and AI Overviews. The problem isn't the advice. It's that the study revealed a 30-point gap most practitioners are not accounting for, and it changes which pages are worth refreshing.

The finding most content guidance skips

Seer Interactive's July 2026 study extended their earlier citation research program -- the same group that produced the widely-cited "79% of AI citations target content under two years old" figure. This time they split freshly updated pages by publication age and measured citation rate for each group.

The result:

- Pages originally published more than one year ago with a fresh update: **72% citation rate** - Pages originally published within the past year with a fresh update: **42% citation rate**

Same freshness signal. Thirty percentage points apart in outcomes. This age-plus-freshness interaction held across all three platforms in the study. It is not a Perplexity quirk or a ChatGPT-specific behavior.

In our August 2026 analysis of this research, we noted what the mechanism explains: AI retrieval systems are not just reading your updated timestamp. They are reading the update in the context of everything else they know about that page -- its index history, its directory corroboration, its link signals, whether it has been retrieved before. A 2022 services page that you add a 2026 FAQ block to is a different signal than a 2025 blog post with a new paragraph added. The authority track record is what makes the freshness update meaningful.

What the half-life data adds to this picture

In our August 2026 review of citation persistence research (session 103), we tracked four independent datasets that converged on the same underlying behavior.

Trakkr's 10-month study of 108,650 citations found that 73.5% of AI citations are one-time events -- cited once, never retrieved again. Mean active duration was 6.8 days. The 30-day half-life figure from this study aligns with Scrunch and Stacker's earlier 3.5-million-event survival analysis (4.5-week median across platforms; ChatGPT at 3.4 weeks, Perplexity at 5.8 weeks).

MaxAEO's B2B SaaS study found a 3-day median for Perplexity competitive recommendation lists -- how long before a brand rotates out of a "best X for Y" answer. Profound's analysis of 3.25 billion citations across 7 AI models found 40-60% of citations rotate monthly.

GetMentions AI's monitoring of 530,875 sources found that 84% of cited URLs are cited by only one AI engine. Not just different from each other -- cited by exactly one platform and invisible to the others.

The picture that emerges: citation presence is perishable. Most content that gets picked up once doesn't get picked up again. Most content that is cited is only cited by one engine. And the pages that survive multiple refresh cycles are the ones with established authority signals supporting the fresh update.

The 3.2x boost applies to pages that can use it

In our July 2026 research (session 88), we consolidated three independent data points on the freshness multiplier: ConvertMate's analysis of 80 million citations, Kevin Indig's analysis of 1.2 million AI citations, and GrowByData's 2026 benchmark. All three converge on content updated within the last 30 days earning 3.2x more AI citations than older content on average.

That number is real. But it belongs next to the Seer Interactive finding: the 3.2x boost is not uniform. A 3.2x multiplier applied to a 42% baseline citation rate (new page, fresh update) yields a different outcome than the same multiplier applied to a 72% baseline (established page, fresh update). The update is the same action. The starting point is not.

Most SMBs with both old and new pages are treating every page the same when it comes to refresh priority. The research says otherwise: the established pages are the high-leverage freshness targets.

The slow intake, fast outflow problem

Our content freshness analysis has used the phrase "slow intake valve, fast outflow valve" since the Ahrefs and Scrunch/Stacker data was reconciled in May 2026. The two studies appeared to contradict each other: Ahrefs found AI citation ages averaging 1,023 days for ChatGPT (nearly three years), while Scrunch and Stacker found content cycling out of the citation pool in 3-8 weeks. Both are correct. The citation pool is conservative about what it lets in, and aggressive about replacing it once it's inside.

Freshness optimization sits entirely on the outflow side. It helps pages stay in rotation once they are inside the citation pool. It does not help pages that haven't entered the pool yet. A business whose pages have never earned AI citations will not see movement from content updates because there is no established pool position to maintain. Phase 1 -- entity infrastructure, directory presence, NAP consistency -- must come first. Freshness is a Phase 2 lever.

Which pages to target and how

If you have an established web presence with pages that have been live for two or more years, the content update strategy should start with the oldest stale pages, not the newest ones.

High-priority freshness targets: - Service pages published in 2022 or 2023 with no substantive updates since - Location or service-area pages with stable URLs but outdated city-specific details or pricing notes - FAQ pages that predate the current AI search environment (anything from 2024 or earlier that hasn't been revisited)

Low-priority freshness targets (for citation purposes specifically): - Blog posts or content published within the last 6 months - Pages created during a site redesign that don't have index history - New content built purely to add freshness, without an established URL

The update itself should modify both the meta timestamp and the visible on-page content. Changing only the dateModified meta tag without updating visible copy produces a mismatch that retrieval systems can detect. The mechanism requires the visible content to be genuinely newer -- a 2026-dated FAQ block, a recent project outcome, an updated pricing range, a relevant regulatory or code update for trades.

For pages in competitive recommendation categories, the MaxAEO data on 3-day Perplexity turnover sets the urgency: this is not a one-time update task. Staying in Perplexity's recommendation rotation for a competitive service category requires regular fresh content additions, not a single refresh. Multiple independent citation sources also extend persistence -- each directory listing is an independent citation that can re-enter the rotation even as any individual page cycles out.

If you don't know which of your pages are being cited

The freshness strategy above requires knowing which of your pages are in the citation pool -- which pages have established positions worth refreshing. If you don't have that data, the prioritization is guesswork.

Signal Check at sourcepull.ca runs across ChatGPT, Perplexity, Gemini, and Claude and surfaces which pages are currently being cited, which platforms are citing you, and where your citation gaps are by platform. If you're at zero citations across platforms, the freshness update strategy is premature -- that's a Phase 1 infrastructure problem, not a content problem. If you're in the pool on some platforms and not others, the page-level citation data tells you where the established positions are and which direction to focus the update effort.