If you've been following GEO advice at all, you've probably heard that large language models love fresh content. You need to update your content constantly, so ChatGPT will cite it. And there is some truth to this, but it's much more nuanced than people typically present it.
You may have heard stats like 70% of the content that ChatGPT sites was updated in the last 12 months. And that is true, but just consider that most of the content people are interested in has probably been updated in the last 12 months. I think it's also important to consider that 30% of all the content ChatGPT sites has not been updated at all in the last year.
Also, how does a large language model know that content is fresh? Let's talk about that and when you should actually care. My friend Dave Davies recently shared something really interesting in Marie Haynes search bar.
By inspecting the page source code while using ChatGPT. He found that ChatGPT actually passes a relevancy parameter to the web tool whenever it performs a search. What this means is that for some searches, ChatGPT will limit the results to pages that have been updated or published within a certain time frame.
So how does ChatGPT know when a page has been published or updated? It just looks for a date on the page, at least right now. It's incredibly easy to trick ChatGPT.
Just put the year in the title or a published date somewhere. At the top of the article. Google is a bit more sophisticated.
It typically won't re evaluate a page unless you make some changes to the content. However, we found that just adding or changing a few sentences is usually enough, so whenever it makes sense. I do recommend updating important pieces of content once per year, including the year in the title.
For your most important pages, you might consi
se. I do recommend updating important pieces of content once per year, including the year in the title. For your most important pages, you might consider updating them every three months.
However, don't fall into the trap of thinking that your pages need to be recently updated to be cited by large language models. Some information really just doesn't change that often, and there's plenty of pages that are years old to get cited frequently. One of the biggest misnomers is Reddit.
When people see how frequently large language models site Reddit, they typically start logging onto Reddit and posting on recent threads. And there's nothing wrong with this. I think being active on Reddit is worth it for its own sake.
But if you look at the threads that large language models are actually citing from Reddit, almost all of them are more than six months old, and most of them are more than one year old. You'll have a much bigger impact on a large English model by posting to the 20 most cited threads than you will by posting 2,000 comments on brand new threads. This was the No.
7 AI. Search factor from Kevin Indigo's growth memo. I recommend going back and watching number one.
We have four more to go.
