The Average Page ChatGPT Cites Is 500 Days Old

A study of 1.4 million ChatGPT prompts, using February data from the ChatGPT 5.2 desktop client, measured the age of every page the model retrieved and compared cited pages against non-cited ones. Within the search retrieval channel, where 88% of all ChatGPT citations originate, the median cited page was roughly 500 days old. That is about a year and four months. Non-cited pages from the same channel were significantly younger.

Freshness, which has been a ranking signal in traditional search for over a decade, does not appear to help a page earn a ChatGPT citation from the search channel. Older pages get cited more often. The relationship runs in the opposite direction from what most SEO instincts would predict.

But the story changes completely in the news channel. Cited news pages have a median age of about 200 days. Non-cited news pages trend closer to 300 days. For news content, recency helps. For search content, age helps. ChatGPT applies different logic depending on which retrieval channel a page enters through, and that distinction has real consequences for how content gets planned, published, and maintained.

Older search pages get cited because they have had time to earn trust

The 500-day median is not a coincidence or a quirk of the dataset. A page that has been live for a year and a half has had time to accumulate backlinks, earn editorial mentions, get referenced across third-party publications, and build the kind of cross-source consensus that language models lean on when deciding which page deserves a citation.

A page published last week might have better information, a cleaner layout, and a more current perspective. None of that matters to the citation decision if the page has not yet built the external validation that signals trustworthiness. ChatGPT is not evaluating which page has the best content in some absolute sense. It is evaluating which page the broader web treats as authoritative, and that kind of authority takes time to build.

The pattern aligns with how the search channel works. 88% of ChatGPT citations come from the general search index, meaning the model retrieves pages based on the same ranking signals Google uses. Pages that rank well in Google tend to be pages that have accumulated backlinks, domain authority, and topical relevance over time. ChatGPT inherits that age bias from the search index it retrieves from.

For anyone producing content with AI visibility in mind, the implication runs against the “publish and move on” approach that works for social media or news cycles. A page designed to earn ChatGPT citations from the search channel needs to be built for durability. It needs a topic that will remain relevant for at least a year, a structure that stays useful without constant updates, and a plan for building external references that compound over the months after publication.

The news channel rewards recency, but only when relevance scores are close

The news ref_type inside ChatGPT behaves differently from the search ref_type. Cited news pages have a median age of about 200 days, while non-cited news pages are closer to 300 days. Freshness helps in this channel, but with an important qualifier.

The study measured title-to-fanout similarity (how closely a page title matches the specific sub-questions ChatGPT generates internally) for both cited and non-cited news pages. The scores were nearly identical. Unlike the search channel, where cited pages showed significantly higher title relevance than non-cited ones, the news channel showed almost no relevance gap between cited and non-cited pages.

When two news pages are equally relevant to the sub-question being asked, ChatGPT appears to use recency as a tiebreaker. The page published more recently wins. But recency alone does not override a relevance advantage. A news page from six months ago with a title that precisely matches the fanout query will still beat a page from yesterday with a vague title.

The two-channel dynamic creates a strategic fork. Content aimed at the search channel benefits from longevity and accumulated authority. Content aimed at the news channel benefits from speed and publication timing. The same page cannot optimize for both simultaneously, because the age signals work in opposite directions.

Which channel a page enters through is not always obvious

ChatGPT assigns each retrieved page to a ref_type internally: search, news, reddit, youtube, or academia. The assignment is not based on what the publisher intends the content to be. It is based on how ChatGPT’s retrieval system categorizes the source.

A blog post on a news publisher’s domain might enter through the news channel. An analysis piece on the same domain might enter through the search channel. A company blog post on a corporate domain will almost certainly enter through the search channel regardless of how timely the topic is. The channel assignment determines which age logic applies, and the publisher does not always control which channel the page ends up in.

The practical test is to look at where a domain’s pages typically appear. Pages from recognized news outlets (Reuters, BBC, trade publications with regular news output) tend to enter through the news channel. Pages from non-news domains (company blogs, comparison sites, educational resources, e-commerce pages) almost always enter through the search channel. Pages on domains that publish both news and evergreen content can split between channels depending on the individual page.

Understanding which channel a page is likely to enter through determines whether to optimize for longevity or for speed, and building that assessment into the content calendar before publication saves effort that would otherwise go into updating pages that were never going to benefit from freshness in the first place.

The compounding value of pages that age well

If the median cited search page is 500 days old, then pages built to last are earning citations long after they were published. A comparison guide published a year ago that still ranks, still carries accurate information, and has accumulated backlinks over twelve months is performing better in ChatGPT citations than a newer version of the same guide that has not yet built comparable external validation.

The compounding works through a feedback loop. A page published on an authoritative domain accumulates backlinks over time. Those backlinks improve its position in the search index. A stronger position in the search index increases the page’s chance of being retrieved by ChatGPT. Being retrieved and cited by ChatGPT can generate additional references and traffic, which feeds back into the page’s authority. Each month the page stays live and continues earning external references, it becomes more likely to be cited again.

Link building accelerates the early phase of this loop. A new page with no external references has to wait for organic link acquisition, which can take months. Building backlinks from authoritative sources compresses the timeline from publication to citation eligibility. The 500-day median represents pages that went through this accumulation process organically. A deliberate link building program moves a page into the same authority range faster.

Digital PR adds a specific kind of reference that carries extra weight. An editorial mention in a trade publication or a tier-one news outlet does not just add a backlink. It adds a citation from a source that ChatGPT’s retrieval system already trusts through the news channel. The page being referenced gains authority in the search channel while the referring article contributes its own credibility through the news channel. The two channels reinforce each other.

Freshness still matters, but for a different reason than expected

The age data does not mean content should never be updated. It means that freshness signals work differently in ChatGPT than in traditional search. Google has long used freshness as a ranking signal for queries where recent information is likely to be better (product releases, current events, pricing changes). ChatGPT appears to weigh freshness much less in the search channel and much more in the news channel.

Updating a page to keep the information accurate remains important for user experience and for maintaining the page’s ranking in Google, which in turn maintains its position in ChatGPT’s retrieval pool. But updating a page purely to make it look fresher, without improving the substance, does not appear to move the citation needle in the search channel. The 500-day median suggests ChatGPT is comfortable citing pages that are over a year old, as long as the external trust signals remain strong.

Where freshness does matter is in the news channel for time-sensitive topics. A brand publishing original research, survey data, or industry analysis timed to a current event can earn news-channel citations that benefit from recency. The key is understanding that these pieces play by different rules than evergreen content. They need to be published quickly, titled precisely to match the sub-questions ChatGPT will generate around the event, and distributed through channels that ChatGPT’s news ref_type monitors.

Two age strategies for two retrieval channels

The age data from 1.4 million prompts splits content strategy into two lanes that require different approaches.

For the search channel, where 88% of citations originate and older pages perform better, the strategy is to build durable pages on topics with long relevance windows. Comparison guides, methodology explanations, detailed product analyses, and evergreen educational content all fit this lane. These pages need a sustained link building and guest posting program to accelerate the authority accumulation that would otherwise take 500 days to develop organically. Each backlink earned in the first six months after publication compresses the timeline to citation eligibility.

Link insertions into existing authoritative pages offer a shortcut specific to the search channel’s age dynamics. If the model favors older pages with established authority, then inserting a brand reference into a page that already has 500 or more days of accumulated trust puts the brand inside a citation-eligible page immediately, rather than waiting for a new page to age into that zone.

For the news channel, where cited pages trend younger and recency breaks ties, the strategy is to publish timely content quickly and distribute it through channels that feed the news ref_type. Original research tied to current events, rapid analysis of industry developments, and data-driven commentary on trending topics all benefit from speed. These pieces have a shorter citation window, but within that window, being first with a precisely-titled page carries a real advantage.

Most content programs already produce both types of content. The age data from this study provides a framework for understanding why some pages earn citations months after publication while others need to land quickly or miss the window entirely. The retrieval channel determines which age logic applies, and the content calendar should reflect that.