I wanna talk about the most effective strategy for being recommended by large language models like Chat, G P T. And Google's A I. Mode.
You've been watching my videos or following other people talking about G E O. You might already have an idea of what I'm gonna say. It's listicles, right?
And that's half true. Listicles are incredibly effective right now. But I'm seeing some data that makes me think it's not for the reason people think.
And if I'm right, this could open up some new opportunities that are equally as impactful and less icky. When someone goes to a large language model looking for a recommendation, the first thing the large language model is going to do is perform a search in a traditional search engine. And the reason Chat, GPT, or Google's A I.
Mode does this is because it's looking for recommendations across the internet, and that's why it loves listicles. These are lists of brands or products, and these pages typically have titles like Best X for Y or best X in Y. And because of this, at least so far, the most effective strategy for being recommended by these large language models is to get yourself mentioned in these listicles.
Or what's typically easier is just to spam these listicles on your own website and any other websites you have access to. So we do listicle outreach for all of our clients to help them get in existing listicles. But if it Makes sense for the client's brand.
We'll also create listicles on their website. However, some new data has me thinking that we actually don't need to create listicles to capture the same benefit. You have been saying for months now that large language models love listicles, but that's not quite true.
The truth is that large language models love to search for things like best X and Y. They search for recomme
ove listicles, but that's not quite true. The truth is that large language models love to search for things like best X and Y. They search for recommendations.
It just happens to be that the vast majority of pages that are optimized for these terms are listicles. And this is because the vast majority of these listicles are automatically generated by directory websites. Once you have a directory and a way to rank different businesses or products, it's very easy to programmatically generate as many listicles as you want, subdividing by the specific product and service along with location, demographic, or use case.
And so, with hundreds of millions of listicles online, it's no wonder that we're commonly seeing large language models citing listicles. How over? We've been tracking thousands of prompts across dozens of brands over the last six months.
And after taking a closer look at that data, I'm starting to think that listicles aren't necessary at all. As I was saying before, it can be kind of icky to put a listicle on your own website, ranking you number one. And for some of our clients, they can't even refer to themselves as the best for legal reasons.
So for clients like this, We've been creating listicles like how to choose the best X in Y, and so far they seem to be doing just as well as listicles. We've also seen a number of our clients competitors that are pumping out hundreds or even thousands of articles like this, and in some cases they're cited more often than the top directories in their space.
Now, it's important to note that the article still needs to explicitly recommend your brand by name, otherwise it's not going to have any impact on the large language model's recommendation. And you want that explicit recommendation as close to the top of the article as
going to have any impact on the large language model's recommendation. And you want that explicit recommendation as close to the top of the article as possible, preferably in the first sentence. Now, that might still feel a little icky to you, but it sure beats putting a listicle on your website with you at the top.
Now, to be completely honest, the brands that are having the most success are doing both the listicle and the how to choose the best article. There's really no limit to the effectiveness of recommendation spam. Now, I should also note that if you're doing thousands of these kinds of posts every month, you are at risk of getting a penalty from Google.
It's called a scaled content abuse penalty, but it has to be pretty egregious before you're at risk. So my advice is to make a 2 by 2 grid in something like Google Sheets on the x axis. You can list all of your products and services on The Y axis.
You can list different demographics or use cases or service areas, whatever makes sense, and then start creating the articles prioritizing the most common combinations.
