In the last two years, your evergreen blog post quietly grew an expiration date.
Nobody sent a memo.
Google’s AI Overviews simply started treating some pages like day old-bread and others like a decent bottle of wine. The ones getting quoted were, more often than not, the ones somebody had recently touched.
Freshness and “last updated” dates influence AI Overview citations, but not quite the way most marketers assume. Recency is a real signal, not a golden ticket. Google’s AI systems weigh how recently a page was substantively revised alongside the authority, structure and how directly the content answers the query at hand.
A five-year-old blog post written by an expert who knows the subject can still beat a page “updated” yesterday with nothing new to say.
What Freshness Actually Means to an AI Overview
Freshness sounds like a simple concept until you try to define it.
Is it the publish date?
The date in the byline?
The lastmod value buried in your XML sitemap?
All three, and Google trusts none of them completely.
For example, a site can quietly rewrite its sitemap timestamp on every server build, which tells Google absolutely nothing about whether a human being actually edited the page.
Google own’s guidance for people-first content draws a line between substantial improvement and cosmetic date-stamping: changing the byline without changing the substance is treated as noise, not a freshness signal.
In this case, the actual mechanism AI systems care about is closer to: has the underlying claim, statistic, or recommendation changed recently enough to still be true? A date is just the shorthand humans use to guess at that.
Another factor is Query Deserves Freshness, a decades-old Google concept that has aged rather well into the AI era. It doesn’t apply evenly. A guide to “best DSLR cameras” needs a freshness pulse. A page explaining how a mortgage amortisation schedule works does not, because the math hasn’t changed since 1850.
How Much Does a Date Actually Matter?
Several studies in 2026 have poked at this question with real citation data, and the pattern holds even when the exact multipliers wobble between methodologies.
- Ahrefs analysed roughly 17 million AI citations and found that URLs cited by AI assistants were, on average, 25.7% fresher than the URLs ranking in Google’s traditional top 10 — a gap of about a year (1,064 days versus 1,432 days average age).
- Seer Interactive’s study of more than 5,000 URLs found that 65% of AI bot crawl hits landed on content published within the past year, climbing to 89% within three years.
- Roughly a third of AI-cited pages were updated within the prior three months, per AirOps’ 2026 State of AI Search research, and pages left untouched past that window tend to lose citation priority on competitive topics first.
- Kevin Indig’s State of AI Search Optimisation 2026 report puts a number on neglect: pages that go a full quarter without an update are roughly three times more likely to lose citations they’d previously held.
Read together, those numbers say something fairly blunt: AI systems lean toward content that looks recently maintained. That being said, “leans toward” is doing a lot of work in that sentence, and the next section explains why.
Last Updated vs Published: Why the Label You Pick Changes Your Clicks
Here’s a wrinkle most publishers never think about until their click-through rate mysteriously drops. Showing both a “Published on” date and an “updated on” date on the same page confuses readers and search systems alike. SEO consultant Abby Gleason documented a case study in which adding a second, older date next to an already visible update date coincided with a sharp CTR decline. Searchers saw the older numbers, assumed the page was stale, and clicked elsewhere.
The fix is almost embarrassingly simple: pick one date, keep it accurate, and stop displaying two competing timestamps that argue with each other. Search Engine Land’s guide to byline dates notes that Google leans on these dates specifically to judge recency on news, reviews, and YMYL (your-money-or-your-life) content, which makes consistency between your visible date, your schema markup, and your sitemap lastmod value worth the ten minutes it takes to fix.
The Catch: Freshness Isn’t a Cheat Code
Here’s the part that should stop anyone from simply bumping every date to “today” and calling it a strategy. A 1,000-query analysis of AI Overview citations found the median cited page was 14 months old, and for evergreen topics a well-maintained two-year-old page routinely beat a fresher, thinner competitor.
Domain authority correlated more strongly with citation odds than age did. Pages over 2,500 words earned 1.6 times the citation rate of pages under 800 words, and pages that named at least one credible source in the body text got cited 2.1 times more often than pages that named none.
A separate meta-analysis by Cyrus Shepard, pulling together 54 studies on AI citation behaviour, ranked the strongest predicators by evidence strength: whether AI crawlers can actually access the URL, whether the page already ranks in classic search, whether the publisher allows content previews, and how directly the page answers the specific question asked.
Freshness didn’t crack the top five. It’s a tiebreaker, not the whole game; win at traditional SEO first, and freshness becomes the thing that decides which of your already-strong pages gets the nod.
Also read:
AI Isn’t Just for Big Brands — Here’s How Small Businesses Are Profiting Too
How to Actually Earn the “Fresh” Label
Let’s say a mid-sized SaaS company has a comparison page “Tool A vs Tool B”. It has been sitting untouched since 2023. Pricing has changed twice. A competitor added a feature that makes half the page inaccurate. The temptation is to open the CM, retype the date, and move on. That’s the move that tanks trust the moment an AI system or human cross-checks a claim against reality.
The better move: rewrite the pricing table with verified current numbers, add a paragraph addressing the new feature, update any screenshots that show an old interface, and only then update the byline date so it matches work that actually happened.
It is worth noting that AI Overviews increasingly draw on structural freshness cues beyond the date itself (visible changelogs, version notes, “what changed” call-outs), because these provide a crawler with corroborating evidence that the update claim is real rather than cosmetic.
A Short Freshness Checklist for Publishers
- Audit high-traffic pages first; a 60–90-day refresh cadence suits commercial pages competing hardest for AI citations.
- Show one date, not two, and keep the visible date, the schema dateModified, and the sitemap in agreement.
- Change substance before the date — new data, new examples, corrected claims — not just the timestamp.
- Keep pages reachable: no aggressive bot-blocking, since AI crawlers can’t cite what they can’t read.
- Skip freshness theatre on stable topics. Constant rewrites can introduce more errors than they fix.
The Takeaway
Freshness dates aren’t magic, and they were never meant to be. They’re one signal among several that AI Overviews use to decide who’s earned a mention, and treating a timestamp as a substitute for actual expertise is the fastest way to build a page that looks fresh and reads hollow.
The publishers winning citations in 2026 tend to be the ones treating “last updated” as a promise rather than a decoration. The date changes because the page got better, not the other way around.
Read:
Frequently Asked Questions
Does changing the date without changing the content help AI Overview citations?
Rarely, and it can backfire. Google treats cosmetic date changes as a quality red flag, not a freshness win.
How often should I update a page to stay competitive for AI citations?
Every 60–90 days for commercial or fast-moving topics. Evergreen explainers can go far longer without losing ground.
Is “last updated” more important than domain authority?
No. 2026 studies consistently rank authority, structure, and answer-relevance above raw recency. Freshness mostly breaks ties.
Should I show both a published date and an updated date?
Not on the same line. Case-study data links dual-date displays to lower click-through rates, so pick one and keep it honest.
Also Read:
