“The content AI cites has a short life. You need to constantly refresh it because LLMs have a recency bias,” said every PR and GEO maestro out there.
And they even have data. Reams and reams of it from big-ass sample sizes.
I am not here to argue with extensive research studies or the intelligentsia behind them.
However, what I know is this: We wrote this sucker on the CES innovation awards last August. I say we because in this case the writers were both human and bot-tacular. It is currently the page with the most agent traffic on our site, if Scrunch is to be believed (the screenshot from our monitoring software looks at the last few hours, but the trend holds over months). It has been in the top 5-10 pages of our site since it was published in August of 2025. It has never been updated since that time.
If took aggregate data as our guide, this article is stale bread. So stale it might actually kill someone with bacteria.
What’s behind this?
It is hard to make a strong causal argument. But we seem to be a recognized authority on CES as far as LLMs are concerned. Other pieces of content including this human-written beast and this wrap piece are regularly scraped and cited (again, despite the fact that they are very old). We also regularly get surfaced in discovery level prompts around agencies that can help companies with their CES presence. We had 3 show clients come from AI last year.
But I am, in this instance, less interested in all that and the reasons why this article and others that seemingly should be put out to pasture are still delivering. I suspect the answers are complex and multifaceted, both things that don’t work nice on the internet or when an expert wants to write a clear-path-to-glory playbook.
Mostly, I want to raise the same thing I raised the other day about listicles. Your data should be your guide in GEO. Not the intelligentsia, not aggregate research. The actual shit you see in your monitoring software.
This should be simple and intuitive. But because experts are clamoring to make generalizable statements and mint their how-tos du jour, you may be pulled towards creating content that doesn’t work for you.
So just ignore the noise and follow the data.