Last updated October 2026.
Two research teams published freshness studies on the same question, four months apart, and they point in opposite directions.
Seer Interactive tracked citation age across ChatGPT, Perplexity, Gemini, and (in an earlier round) Google AI Overviews, and found a strong, consistent bias toward pages updated in the past year. Digital Applied crawled 4,243 URLs cited inside Google AI Overviews and found the opposite: a median page age of 14 months, with no statistical link between recency and citation once domain authority was controlled for.
Both studies used real data. Neither is wrong. The contradiction is the story, and the engine you care about decides which finding applies to you.
The case for fresh content: Seer Interactive’s two studies
Seer Interactive has published two rounds of freshness research, one in June 2025 and a follow-up in July 2026, and both point the same direction.
The first study pulled publish and update dates for more than 5,000 URLs cited across ChatGPT, Perplexity, and Google AI Overviews. The skew toward recent content held on all three engines, but the size of that skew varied:
- ChatGPT: about 31% of citations traced to 2025, with a tail running back to 2004, largely high-authority reference pages such as Wikipedia.
- Perplexity: about 50% of citations traced to 2025 alone.
- Google AI Overviews: about 44% of citations traced to 2025, the strongest recency lean of the three engines tracked that year.
A separate slice of that 2025 study, built from ChatGPT crawler log-file hits rather than citations, found the same lean: 65% of AI bot hits targeted content published within the past year, 89% within three years, and only 6% on content older than six years.
The 2026 follow-up switched to a citation dataset built from actual client data, dropped Google AI Overviews, and added Gemini. Across 7,683 dated pages carrying 47,097 citations for four client brands, 75% of cited pages had been updated within the last year, and 88% within two years. The strongest lean was Gemini at 78%, then ChatGPT at 73%, then Perplexity at 65%.
One nuance matters here: in Seer Interactive’s data, freshness mostly comes from updating old pages, not publishing new ones. More than a quarter of the pages the 2026 study classified as fresh were originally published two or more years earlier. A third source backs the same direction: Ahrefs’ July 2025 analysis of 17 million AI citations found AI-cited content runs 25.7% fresher, on average, than content cited in Google’s traditional organic results.
The case against freshness as a ranking factor: Digital Applied
Digital Applied ran a different kind of study in April 2026. Instead of asking what year cited content traces to, the team asked which page-level signals independently predict a citation once you control for the obvious confounds.
The team sampled 1,000 Google AI Overviews across 10 query intents and about 30 verticals, then crawled the 4,243 unique URLs that showed up in citation lists. Those pages were compared against a control set of roughly 50,000 URLs: the next 50 organic results for the same queries. Every URL was scored on word count, schema markup, last-modified date, named-source citations, and a domain-authority proxy.
The headline finding: the median cited page was 14 months old. Once domain authority was accounted for, recency showed no independent link to citation. Four other signals did: domain authority (+0.61 correlation), schema markup (cited 2.3 times more often), body content over 2,500 words (1.6 times), and at least one named-source citation in the body (2.1 times). The same dataset found the top 1% of cited domains, about 12 sites, capture 47% of all citations, which is part of why the team treated domain authority as the variable to control everything else against. Digital Applied’s own note on the miss: recency mattered only for explicit news-intent queries, which sat outside the 10 intent classes it sampled.
The two studies, side by side
| Study | Sample | Finding | Limitation |
|---|---|---|---|
| Seer Interactive, “AI Brand Visibility and Content Recency” (June 2025) | 5,000+ dated URLs cited by ChatGPT, Perplexity, and Google AI Overviews | Google AI Overviews: ~44% of citations from 2025. Perplexity: ~50% from 2025. ChatGPT: ~31% from 2025, tail to 2004. | Raw citation-age distribution, no domain-authority or query-intent controls; single vendor’s client base. |
| Seer Interactive, “Content Recency’s Impact on AI Visibility in 2026” (July 2026) | 7,683 dated pages, 47,097 citations, ChatGPT, Gemini, and Perplexity, four client brands | 75% of cited pages updated within the last year; 88% within two years; freshness mostly comes from refreshing old pages, not new publishing. | Four brands across four verticals; no Google AI Overviews coverage; dates read from schema, sitemaps, and headers. |
| Digital Applied, “1,000 AI Overviews Analyzed” (April 2026) | 1,000 Google AI Overviews, 10 intents, ~30 verticals; 4,243 cited URLs vs. ~50,000 control URLs | Median cited page age: 14 months. No independent recency correlation once domain authority is controlled for. | Google AI Overviews only; two-week US desktop sample; excludes explicit news-intent queries. |
Why the two studies reach opposite conclusions
Both studies are honest. They are not measuring the same thing.
Seer Interactive’s numbers are a raw citation-age distribution. The team pulled every dated page cited by a set of client brands and asked what year each page traces to. That method is fast and real, but it cannot separate freshness from the things that travel alongside it, such as domain authority and content type.
Digital Applied’s numbers come from a controlled comparison against a same-query control group. Recency lost that test. Schema markup, content length, and named-source citations won it.
Put the two side by side and a pattern appears. Seer Interactive’s own 2026 update found that more than a quarter of “fresh” pages were originally published two or more years earlier. Freshness, in other words, is often manufactured by updating an already-authoritative page, not by publishing something new. That lines up with Digital Applied’s finding: raw page age does not predict citation, because a page’s authority, structure, and sourcing carry more of the signal than the date stamped on it.
A widely repeated industry figure, credited to Amsive’s Lily Ray, adds a sharper edge to the freshness camp: roughly half of all AI citations trace to content published in the last 13 weeks. Treat that number as directional support rather than fully verified, since the underlying methodology has not been published in full. It points the same direction as Seer Interactive’s data, not toward Digital Applied’s.
The honest read: fresh content and cited content cluster inside the same pool of well-maintained, high-authority domains. Freshness alone, isolated from authority, is a weaker lever than either headline number suggests on its own.
The per-engine verdict
Skip the single universal answer. Here is what the data supports, engine by engine.
Neither study breaks out Microsoft Copilot on its own. Until a dataset does, treat Copilot’s freshness sensitivity as unproven and test it directly with the method below rather than assuming it matches any of the four engines above.
How to test freshness decay on your own pages
You do not need to trust either study blindly. Citation-tracking tools let you run the test on your own domain.
Otterly.AI and Profound both log which of your URLs get cited, on which engine, and when. Run this: pick five pages that already earn occasional citations, update three of them with genuinely new information or data, and leave two untouched as a control group. Check citation frequency again in four to six weeks. If the updated pages pull ahead, freshness is a real lever for your niche and engine. If citation rates stay flat, you have a Google AI Overviews-style case on your hands, where authority and structure carry more weight than a last-modified date.
Trigger-rate context matters here too. BrightEdge’s tracking put Google AI Overviews on roughly 48% of monitored queries as of February 2026, so a Google-specific freshness test only tells you something if you are picking queries where an AI Overview actually renders in the first place.
For teams that want tracking, gap diagnosis, and content execution in one subscription instead of stitching several tools together, Temso covers this workflow across ChatGPT, Perplexity, Gemini, Google AI Overviews, and Microsoft Copilot, from $89/mo.
Do not guess your update cadence from a headline stat, yours or anyone else’s. Pick your primary engine, run the freshness test above for six weeks, and let your own citation data set the schedule. For a full side-by-side of the platforms that can run that test for you, see the GEO tool rankings, and check the glossary for definitions of any term used here.