TLDR Marketing ran a piece this week on what an AI search engine wants from a page, and most of it is the kind of advice you nod through: four to six direct sentences near the top, clear headings, tables and lists where they help, an FAQ. Then freshness. Pages less than three months old, the summary said, are reportedly three times more likely to earn citations.
The freshness part wasn’t news. I’ve listened to Cassie Clark’s Found in AI podcast for months, and freshness is the first leg of her FSA framework (freshness, structure, authority), which she published last December and grounds in her own testing across ChatGPT, Perplexity, and Gemini. Her explanation was already familiar to me.
But back to the TLDR claim. “Three times” is a specific claim sounds pretty compelling, but “reportedly” can go a long way in those newsletter summaries so I was cautious about the findings. It turns out that TLDR was summarizing Chris Long’s September 21 post for Nectiv, “The Perfectly Optimized Page for AEO,” an update of the old Moz infographic for the AI era. Nectiv says AI systems are three times more likely to cite pages less than three months old and links to AirOps’ December 2025 report, “The 2026 State of AI Search.”
The problem is that the AirOps report says something else: pages not updated quarterly are more than three times as likely to lose AI citations. Losing a citation you already have and earning one you don’t are different events. A comparison of citation loss among previously cited pages doesn’t establish the chances of getting cited in the first place. Nectiv also shifts from time since an update to the age of the page. By the time the number reached my inbox, the multiplier remained, but what it multiplied had changed. I understand how these things happen when content is shared in multiple contexts that can introduce a change in the record in a very human way. But it turns out that the source wasn’t entirely consistent either.
The December report links to an August AirOps study whose summary says pages untouched for over a year are more than twice as likely to lose citations. Its body switches between losing citations and being less likely to earn them. It also gives fast-moving industries a three-month freshness window. I could find pieces of the later claim there, but not the analysis connecting them to the quarterly, 3x figure.
That study examined more than 4,000 ChatGPT-cited pages across 900 high-intent queries in 15 industries. A sample of cited pages can tell you how old cited pages are. By itself, it can’t tell you how much more likely a fresh page is to be selected than an eligible stale one. The public methodology also doesn’t describe tracking citation loss over time. And the page gives two different figures for the share untouched for over a year: 26.2 percent in one section and 56.4 percent in another.
AirOps sells tools for refreshing content. That doesn’t invalidate its research, but it makes the missing explanation worth asking for. What I can document is narrower than a clean story about a statistic getting worse at every stop. AirOps published different thresholds and multipliers in two reports, with inconsistent language in the earlier one. Nectiv changed the meaning of the later report’s headline, and TLDR repeated Nectiv’s version. I couldn’t find enough public analysis to check the quarterly multiplier itself.
Other versions circulate too. Status Labs says content updated within 30 days receives 3.2 times more citations than older material. I couldn’t find transparent support for that exact number on the page either. Meanwhile, The Digital Back Office’s freshness article retained AirOps’ distinction about losing citations. Repeating the source accurately was possible.
Before writing this, I ran the TLDR claim through Perplexity for a fact-check. It caught the difference between losing and earning citations. It also described the December report’s quarterly, 3x claim and attached a footnote linking to the August report, the one whose summary says twelve months and 2x. Its follow-up source audit separated the reports correctly. The first report had supplied a useful correction with the wrong evidence attached. That’s plausible fiction at footnote scale, inside an audit of someone else’s sourcing. Trust but verify won’t be going away anytime soon.
There are still reasons to take freshness seriously. Ahrefs analyzed roughly 17 million cited URLs and found that AI-cited pages were about 26 percent newer than organic Google results by publication date and about 13 percent newer by last update. The pattern varied across platforms: ChatGPT cited newer material, while Google’s AI Overviews looked much closer to ordinary Google results. The average AI-cited page was still about 2.9 years old by publication date. That’s a useful detail to keep beside a three-month deadline but don’t take that that as permission for complacency.
SE Ranking found that pages updated within three months averaged 6.0 ChatGPT citations, compared with 3.6 for older content. That’s a difference in average citation count. It doesn’t tell me what would happen if I opened one of my pages today, revised it, and ran the same questions tomorrow. The study was observational, and the authors acknowledge that the factors they examined depend on one another.
My take-home advice lands close to where TLDR’s did. If the facts on a page can change (prices, services, what you offer this year), keep them current. When you make a substantive revision, put an honest update date on it. What I wouldn’t suggest is rewriting content just for the sake of freshness. You have better things to do with your life.
That leaves the question I’m more interested in now. Cassie’s phrasing includes a useful qualification: the models she tested appeared to prioritize content that appeared recently updated. I control the visible update date on my pages and the dateModified field in their markup. I can change either without correcting a fact. I still don’t know how much that will change the answer.