Do AI engines prefer fresh pages? What the freshness studies actually show
AI assistants cite somewhat newer pages than Google ranks, but still years-old ones. When a real update helps, and why a new date alone does not.
The short answer
Somewhat, but less than the advice suggests. The best-known study found AI assistants cite pages that are newer on average than Google's organic results, yet still typically years old. Update a page when its facts change, and change its date only then. A new date on unchanged content is fake freshness and risks your credibility.
Key takeaways
- AI assistants lean towards newer pages than organic search, but the pages they cite are typically years old, so age alone rarely decides a citation.
- Update a page when its facts change: prices, versions, statistics, regulations or anything a reader would act on.
- Google's guidance says a page's dates must describe its own publication or update, which rules out bumping dates on unchanged content.
- Seer Interactive found pages cited in an AI Overview earned about 120% more organic clicks per impression across 53 brands, January 2025 to February 2026.
- Track the questions where recency matters, such as prices and year-specific comparisons, and leave evergreen pages alone until something changes.
In this article
The short answer: fresher, not fresh
AI assistants lean towards newer pages, but they do not demand new ones. The study most often quoted on this, from Ahrefs, compared the age of pages cited by AI assistants with the age of pages in Google's organic results. Cited pages were newer on average, and still old by any normal sense of the word. We cover its figures below, with their caveats.
That gap explains the advice now circulating: refresh every page every quarter, change the "last updated" date, add the current year to titles. Some of it helps. Much of it is noise that costs time and can cost trust.
Our view: freshness is a property of facts, not of pages. A page whose facts are current is fresh, whatever its date. A page with a new date and last year's prices is stale and misleading at once.
Why the date question matters at all
It matters because a citation in an AI answer is worth having. Seer Interactive tracked 53 brands from January 2025 to February 2026 and found that brands cited in a Google AI Overview earned about 120% more organic clicks per impression than when they were not cited.1 That is a correlation across the brands Seer tracked, not proof that the citation caused the clicks. It is still a strong reason to understand what gets a page cited.
Seer also found AI Overviews on 95.4% of "X vs Y" comparison queries in the same data.1 Comparisons are exactly the pages that go stale fastest, because the products in them change. So the freshness question is sharpest where the clicks are.
What Google says about dates
Google's own guidance is clear about what a date means. Its Search Central page on byline dates says the dates must describe the publication or update date of the page, not the events described in it.2 It asks for a visible date, matching datePublished and dateModified values in structured data, and consistency between the two.
That guidance does not reward a new date by itself. A date is a claim about the page. If the claim is that the page was updated, the page should have changed in a way a reader would notice. Our own posts follow the same rule: updated_at changes only when the substance changes.
What the Ahrefs freshness study found
Ahrefs extracted 16.975 million cited URLs from ChatGPT, Perplexity, Gemini, Copilot, AI Overviews and Google's organic results, published in July 2025 with the data period not stated.3 The study is now a year old, which is a long time in this field, so read it as direction rather than a current measurement.
Pages cited by the four AI assistants averaged 1,064 days since publication, against 1,432 days for organic results, in a study whose data period was not stated.3 Ahrefs called that 25.7% fresher. Measured by last update instead, the gap was smaller: 909 days against 1,047, about 13% (data period not stated).3
Average age of cited pages, days since publication
Two details matter more than the averages. Google's AI Overviews cited pages about as old as the organic results, 16 days older on average in Ahrefs' data (period not stated).3 And Perplexity and ChatGPT appeared to order their in-text references from newest to oldest, which suggests recency affects position within an answer more than inclusion.
The method has limits worth naming. Ahrefs dated a page by when its crawler first saw it, which can be later than the true publication date. Averages hide a long tail: a few very old reference pages can pull the mean up. And Ahrefs itself warned that frequent updates without content changes risk backfiring, and that freshness is one factor among many.
When does a real update help?
A real update helps when the question itself has a time in it. Prices, product versions, regulations, statistics and "best X in 2026" comparisons all change, and an engine choosing between two pages on such a question has reason to prefer the one with current facts.
It helps much less on evidence-based or evergreen pages: how a process works, what a term means, how to choose. Those pages earn citations by being clear and specific, which is the same thing our review of schema, llms.txt and Reddit found for other tactics. Ranking still matters too, as our note on top-10 rankings and AI Overviews explains.
| Change to the page | Update the visible date? |
|---|---|
| Prices, rates or fees changed | Yes, and say what changed |
| New statistic replaces an older one | Yes, with the new source |
| Product version or feature list changed | Yes |
| Typo fixed, image swapped, layout changed | No |
| Year added to the title, nothing else | No, and remove the year |
Our editorial rule. Illustrative; adapt it to your own content.
What is fake freshness?
Fake freshness is any signal of recency that the content does not support. The common forms are a "last updated" date bumped by a plugin on every save, a current year inserted into titles and headings by a template, and a rewritten intro on an otherwise unchanged page.
All three are cheap, and all three can be checked. A reader who sees "updated this month" above a price that changed last year stops trusting the page. Google's guidance treats the date as a statement about the page. Engines that read the page see that the facts did not move.
Our view: if your CMS changes dateModified on every save, turn that off. Set the date by hand when the substance changes, and keep a short changelog on pages where readers act on the facts.
A freshness routine that holds up
A simple routine works better than a blanket refresh. List the pages that answer time-sensitive questions: pricing, comparisons, product specifications, regulation and anything with a statistic. Review those on a fixed cycle, change what is out of date, and update the date only when you do.
Then measure whether it made a difference. Track a fixed set of prompts that those pages answer, and run each on several days, because single runs vary. At Sigzen AI, our AI answer monitor covers six engines (ChatGPT, Gemini, Perplexity, Claude, Copilot and Google AI), and audits run each prompt on three separate days. If citations do not move after a real update, the page has a problem other than its age.
Sources
- Seer Interactive, AI Overviews and CTR, 2026 update: 53 brands, Jan 2025–Feb 2026 (Apr 2026)
- Google Search Central, influence your byline dates (last updated Dec 2025)
- Ahrefs, do AI assistants prefer fresh content: 16.975M cited URLs, five AI surfaces plus organic results (Jul 2025); data period and geography not stated
