You've probably heard that ChatGPT, Gemini and Perplexity like fresh content. That's true, but it leaves out the more useful part. A July 2026 study from Seer Interactive dated thousands of pages that AI assistants cite. It found that 75% of LLM-cited pages were updated in the last year and 88% within two years — but measuring by original publish date instead of update date, only 42% were actually written recently. So the "fresh" content AI engines reward is mostly older pages that someone kept up to date. That changes how a small team should spend its writing hours.

What the Seer study actually measured

This study is worth slowing down for because the method shapes the conclusion. Seer looked at 7,683 pages carrying 47,097 citations across ChatGPT, Gemini and Perplexity, drawn from non-branded answers for four brands in pet retail, vacation rentals, energy and commercial banking, between March and June 2026. Last-modified dates were pulled from schema, sitemaps and headers, and Seer is upfront that only about two-thirds of pages could be dated successfully.

Two limits stand out. First, the study covered four industries, not the whole web. Second, it didn't include Google: Seer's separate study did not cover AI Overviews. Treat it as strong evidence about how chat assistants behave, not as a universal rule.

It also updates Seer's own earlier work. In our original study, in June of 2025, 65% of log hits were for content published within the past year. That 65% figure spread widely and led many people to conclude that new posts win. The 2026 follow-up suggests a different conclusion.

Fresh by update date vs. fresh by publish date

The key finding comes from measuring the same pages two ways. Seer's study measured last-update dates and found many cited pages were maintained older pages, with 72% looking fresh by update date versus 42% published in the last year. (Write-ups of the study quote both 72% and 75% for the update-date figure, depending on the cut. Either way, there's a gap of roughly 30 points.)

Put simply, much of the freshness that AI assistants reward is the product of maintenance on older URLs. Another summary put it more bluntly: "Old URLs win. Stale URLs lose."

The pages that held up longest point the same way. Pages cited consistently across all four months of the study skewed 68% fresh with a median age around six months. The pages that keep getting cited are the ones that keep getting looked after.

Why would an old URL that's been refreshed beat a brand-new page? Seer didn't test the cause, but the likely reason is simple. An older URL has already built up links, mentions and a crawl history, and the update tells retrieval systems it's still accurate. A new page starts with none of that history.

How much freshness matters depends on the engine

Seer isn't the only source. Ahrefs looked at roughly 17 million citations and found that AI-cited URLs were on average 25.7% fresher than the URLs that rank in top-10 organic Google results (1,064-day average for AI vs 1,432-day average for organic).

A 1,064-day average works out to almost three years. In that same dataset, AI-cited pages averaged about 2.9 years old, and Google AI Overview citations averaged closer to 3.9 years. AI engines lean toward recent content, but they still cite plenty of established pages.

Engines also differ from each other. Pages cited in the top three Google AI Overviews were slightly older than Google's organic results, while ChatGPT, Gemini and Copilot cited newer pages. Google's systems appear the least influenced by recency of the platforms studied. If AI Overviews matter most to you, ranking signals still carry more weight than recency. If ChatGPT and Perplexity matter most, keeping pages current counts for more.

Topic matters too. In Seer's industry breakdown, in pet retail, breadth of content wins, recency matters the least. In travel, where prices and availability change all the time, the gap between update date and publish date was among the widest.

Freshness numbers to be skeptical of

Because freshness is a popular topic, some of the statistics you'll see don't hold up. One analysis that traced figures back to their sources warned that the viral '4.3x more cited for fresh content' figure traces to no primary study and should not be repeated.

You'll also see "the 13-week rule," which says half of all AI citations go to content under three months old. It comes from an Amsive analysis, and Amsive's 2026 citation freshness analysis found that 50% of AI search citations come from content less than 13 weeks old. It's a real data point, but it measures one dataset at one point in time. It isn't a deadline after which a page stops counting. Meanwhile, a separate analysis of 1,000 Google AI Overviews, published in April 2026, found a median cited page age of 14 months. Citation slots went to structured, well-sourced content from credentialed domains, more or less regardless of publish date.

When freshness studies disagree this much, look at how each one defines "fresh." It might mean publish date, last-modified date, or date of the bot visit. Most of the disagreement comes from those different definitions.

What a real refresh looks like

The practical lesson is to spend less time producing new posts and more time keeping your best existing posts current. Here's what that looks like:

  • Start with pages that are already cited or ranking. They're the old URLs most likely to benefit from an update. Use GA4 referral data or Search Console to find pages that get AI or organic traffic but haven't been touched in a year.
  • Change the substance, not just the date. Replace outdated statistics, update prices and product names, add what's changed, and delete advice that no longer applies. Changing the timestamp on a page whose content hasn't changed makes it look current when it isn't, and that's the stale content readers and engines are both moving away from.
  • Make sure your dates are readable. Seer pulled update dates from schema, sitemaps and HTTP headers. If your CMS doesn't output an accurate `dateModified` or sitemap `lastmod`, a real update may not show up as one.
  • Match your update schedule to how fast the topic changes. Pricing pages, tool comparisons and "best of" lists go out of date in months. Evergreen explainers don't need the same attention.
  • Keep pages easy to quote. Freshness gets a page looked at, and structure is what gets it cited. Our guide on making content easy for ChatGPT and Google AI to cite covers formatting.

One warning: one client saw +300% AI traffic after refreshing outdated content, but the gain was concentrated in categories where freshness was part of the query intent. Local services and evergreen reference content barely moved. Refreshing pages works best where readers actually need current information.

FAQ

Does changing the "last updated" date help a page get cited by AI?

There's no good evidence that changing the date alone helps, and it can mislead readers. The Seer study tracked dates from schema, sitemaps and headers, so accurate dates matter. But the advantage it found belonged to pages that people actually maintained. Update the content first, then make sure the date reflects that.

Should I stop publishing new content?

No. But don't assume new posts are the main way to get cited. In Seer's data, only 42% of cited pages were published within the past year, while about three-quarters had been updated in that time. For most small teams, a mix that favors refreshing high-value existing pages over publishing more new ones fits the evidence better.

Does freshness matter as much for Google AI Overviews as for ChatGPT?

Apparently not. Ahrefs found Google's AI Overviews were the least influenced by recency of the platforms it studied, and their cited pages averaged close to four years old. ChatGPT, Gemini and Copilot leaned newer. If you're aiming for AI Overviews, normal SEO strength still matters most. Freshness helps more with the chat assistants.