AI citation tracking

If you want your company or brand to be recommended by large language models, you need to be tracking the pages and websites they cite. In other words, you need to be simulating the kinds of prompts that your ideal customer might be typing into these large language models.

And then pay attention to what pages online the large language models look at before they recommend you or your competitors. This is really the entire game right now.

If those pages cited by the large language models recommend your brand, the large language model will recommend your brand. We've been tracking this for our clients for about six months now.

We currently simulate about 4,000 prompts every day across chat, G, b, t, Gemini, and perplexity. We use software called P dot AI to do this, which I highly recommend.

I was actually just on a call with someone from peak yesterday, and he was telling me that a lot of people are wondering what to do with all this information...

to vary wildly by industry. So any advice about, hey, ChatGPT really likes this website or hey, you should be on this website cause Google sites it all the time.

None of that advice is generally true across industries. You really need to be tracking prompts of people looking for the type of service or product that you offer.

Let's talk about directories. Typically, there's gonna be between one and 5 directories that large language models often refer to when people are asking about your industry.

Might be other directories that are referenced 1 in 1,000 prompts, 1 in 10,000 prompts, but there's typically going to be a few that are referenced between 1 and 50% of the time someone asks about your industry. You might hear people saying you have to be in all the business directories cause large language models are looking there.

No, not really. Find the 1 to 5 that actually matter and get your business there.

Next, let's talk about listicles. Large language models love to cite listicles...

ng. Right now outside of directories, the other listicles you can be on might be on a blog post, a LinkedIn article, a medium article.

Typically it's going to be very difficult to get your brand added to these listicles. A lot of them are going to be created by your competitors.

However, it also takes about 30 seconds to send an email and ask what it takes to get on their listicle. So I think it's worth the effort.

In our experience, we have about a 5% success rate. Most of the time, though, they're gonna ask for money or they're gonna ask for a reciprocal link.

However, if it's a page that's often cited by large language models, it's probably worth it. The next one you'll see a lot is discussion threads.

Typically this is gonna be Reddit or Facebook groups. The thing to point out here is it's typically the same five to 10 threads that are being cited.

Over and over again. So we might identify 500 threads that have been cited, but 80% of the citations will go to those top 10.

And these are typically threads that are at least a year old...

xplicitly recommend you and when the large language model cites you, it's going to retrieve a block of text and if in that text isn't the recommendation for your brand, it actually doesn't matter that it cited your website, it's not gonna recommend you in the response. So those are all the ways that you can influence the content that's on the pages large language models are already citing.

However, in our experience, we have a lot more success influencing these responses by creating content ourselves. I feel like most of the advice online right now centers around influencing the content on the Pages that our senior scholars are already citing and they completely gloss over the fact that you can also replace the pages large language models are citing.

And that's typically much easier to do...

There's a lot of conflicting information out there about how to rank in ChatGPT. Just this week I've seen people say that ChatGPT hates press releases.

Well, in my experience, they love press releases. I've heard that ChatGPT never cites Reddit and that ChatGPT is always citing Reddit.

I heard another person I respect saying that large language models don't like LinkedIn, where in my experience, they love LinkedIn. Who's right?

Well, they're all right. What I think is not being talked about enough right now is that the websites that these large and small site depends very much on the prompt itself.

So we're currently tracking about 4,000 different prompts cross 25 or so agencies. We run all these prompts every day across chat, GPT, perplexity, and Google's AI.

Mode. Whenever someone asks me which websites these large language models cite the most often, the only coherent thing I can say is it depends on the industry.

So my advice is to ignore any information you're hearing about ChatGPT loving this website or hating that website...