Half of the content cited in AI answers is under 13 weeks old. Here is what content decay means for your AI visibility, and why a refresh habit wins.
Last updated: 26 June 2026
There is a single statistic that changes how you should think about content and Artificial Intelligence (AI) search. It is the strongest argument we know of for treating AI visibility as an ongoing habit rather than a one-off project. Here it is, what it means, and what to actually do about it.
Across recent studies of which sources Large Language Models (LLMs) reach for when they answer a question, one finding keeps appearing:
Roughly half of the content cited in AI answers is less than 13 weeks old.
Thirteen weeks is one quarter. So half of everything ChatGPT, Claude, Gemini, and Perplexity quote back to a user was published or meaningfully updated in the last three months.
That is a very different world from classic Google search, where a strong page could sit at the top of the results for years and keep earning traffic on the strength of its backlinks alone. In AI answers, freshness is not a tie-breaker. It is one of the main events.
Content decay is the slow loss of visibility a page suffers as it ages, even when nothing about the page itself changes.
In Generative Engine Optimisation (GEO), the decay is sharper than it is in traditional Search Engine Optimisation (SEO). A page that was being cited in AI answers in March can quietly stop being cited by June, not because it got worse, but because newer pages on the same topic arrived and the models started reaching for those instead.
Three things drive this:
First, the models re-evaluate sources on their refresh cycles. Each engine pulls in new material on its own schedule, and when it does, recent content gets a look-in that older content does not.
Second, AI answers prize "current" over "comprehensive" for a large share of queries. If someone asks about wait times, prices, rules, or anything that changes, the model has learned that a recent source is safer to quote than an old one.
Third, your competitors are publishing. Every fresh, well-structured page a competitor ships on your topic is a page that can displace yours in the citation slot, regardless of how authoritative your original was.
The practical takeaway is uncomfortable but useful: in AI search, standing still is the same as sliding backwards.
It would be easy to file "freshness matters" under nice-to-know. The reason it is a bigger deal than that comes down to where AI search is heading, and how the people who arrive through it behave.
AI answers are now a serious share of search. AI Overviews, the AI-generated summaries Google shows at the top of results, now appear on around 25% of searches. That is a quarter of all queries where an AI summary, not a list of blue links, is the first thing the user sees. If your content is not fresh enough to be cited in that summary, you are not in the most valuable space on the page.
The visitors who do click through are worth far more. Analysis from Ahrefs found that visitors arriving from AI search convert at a dramatically higher rate than visitors from traditional organic search, on the order of a 24 to 1 difference in some samples. The logic is intuitive: by the time the AI has answered the surface question and the user still clicks your link, they are deep into a real decision, not just browsing. Fewer visitors, but each one is closer to buying.
The format levers are real and additive. The same body of GEO research that surfaced the 13-week finding also found that adding statistics to a page can lift its visibility in AI answers by around 41%, and that comparison-style articles punch well above their weight, taking roughly 32.5% of citations in the categories where they appear. Freshness gets you considered. Statistics and comparison structure get you chosen.
Put those together and the picture is clear. AI search is a growing slice of all search, it sends you a smaller number of much higher-value visitors, and the content that wins is recent, evidence-backed, and well structured. Freshness is the entry ticket to all of it.
Most businesses, when they first take AI visibility seriously, treat it as a project. They commission an audit, rewrite a batch of pages, add some statistics and structured data, and then move on. For a few weeks, citations climb. Everyone is pleased.
Then the 13-week clock starts ticking. The freshly-rewritten pages age. Competitors publish. The next model refresh reaches for newer material. By the next quarter, the gains have quietly eroded, and the business is left wondering why the AI visibility "stopped working".
It did not stop working. It decayed, exactly as the data says it will. The one-off project was always going to decay, because the thing it was fighting, content age, never stops advancing.
This is the single most important shift in how to think about GEO. It is not a renovation you do once. It is a garden you tend. The businesses that win in AI answers are not the ones with the best single audit. They are the ones with a refresh habit.
You do not need to rewrite your whole site every quarter. You need a small, repeatable cycle aimed at the pages that matter.
Start by knowing which of your pages are currently cited, and which have slipped. You cannot tend what you cannot see, so the first move is measurement: track which of your pages appear in AI answers, and watch for the ones that fade.
Then run a light quarterly pass on your priority pages. Update the dates and any figures that have moved. Add one or two current statistics with their sources. Tighten the structure into short, scannable paragraphs. Add a comparison element where it fits, because comparison content earns more than its share of citations. This is 30 minutes a page, not a rebuild.
Finally, keep shipping new pages on live topics. A steady trickle of fresh, well-structured content does two jobs at once. It gives the models recent material to cite, and it keeps you in the citation slot before a competitor takes it.
The maths is gentle. A business that refreshes a handful of priority pages each quarter and ships a couple of new ones a month spends a few hours a month and stays in the half of content that is under 13 weeks old. The business that does nothing slides out of it, slowly, and then all at once.
A few things to keep in mind so you calibrate this correctly.
The 13-week figure is a pattern across studies, not a hard cut-off. A genuinely authoritative page can keep earning citations well past three months. Freshness is a strong signal, not the only one. Backlinks, domain trust, and clear structure still matter and still compound.
Refreshing for the sake of changing a date does not work. The models are looking for content that is actually current, meaning the facts, figures, and framing are up to date, not just the timestamp. A cosmetic date change with stale content underneath will not earn you the freshness benefit.
And freshness is the entry ticket, not the whole game. Get cited by being recent, get chosen by being useful, evidence-backed, and well structured.
In AI search, content has a half-life measured in weeks, not years. Half of what gets cited is under 13 weeks old. That makes AI visibility an ongoing habit, not a one-off project, and it is exactly why a steady refresh cycle beats the best single audit you could ever commission.
Sources: GEO content-decay and freshness findings summarised in 2026 generative engine optimisation roundups by Omnibound, Superlines, and Arfadia. AI Overviews share of searches and AI-search conversion data via Ahrefs. Statistics and comparison-content citation findings from 2026 GEO research write-ups. Primary sources linked where the roundups provide them.
WhatsGEO is the AI visibility platform from Creative Sauce Ltd. We help businesses measure and improve how often their domain is cited in AI engine answers like ChatGPT, Claude, Gemini, and Perplexity. Measuring is the first step, but the lasting results come from a refresh habit, which is exactly what our agents are built to run for you.