We know from an Ahrefs study that the number of times your brand is mentioned around the web is one of the factors that correlates most strongly with getting your brand mentioned in AI responses.
But this doesn’t capture the nuance of what getting a “good” brand mention on the web looks like.
LLMs associate your brand with specific concepts based on how often your brand appears near those concepts in LLM training data.
For example, “Hubspot” and “CRM” appear in proximity to each other a ridiculous number of times around the web, and subsequently all LLMs “know” Hubspot should be in the conversation when CRMs come up.
Therefore, getting your brand mentioned on the web alongside clear information about your brand and its products is a win for AI search optimization.
Let’s also posit that it works in your favor to get these mentions on reputable websites that are relevant to your industry. Even if ChatGPT isn’t very discerning today, there are signs it will be soon.
The big caveat is that these wins can take a while to pay off. LLMs “learn” from new web mentions of your brand in periodic retraining, which can take months+ to complete.
The get-me-results-faster approach
No one wants to wait months or years to see efforts bear fruit.
To influence AI responses sooner, get your brand mentioned in the content that AI models cite when they do real-time web searches to respond to user queries.
“Well, yes,” you say. “This is known.”
Stick with me.
When you’re tracking an AI prompt that’s really important to your business because it’s so relevant to your offering, you might be tempted to target the URLs that ChatGPT (or whichever AI model) cites when you run that prompt.
Here’s the danger of that approach:
Of the URLs that ChatGPT cites for a given prompt today, less than 30% of those URLs will still be cited a month from now.
To put it mildly: Getting your brand mentioned in content that ChatGPT cites today but doesn’t cite next month is not a sustainable way to improve your brand’s AI visibility. So don’t look at the citations of a single prompt.
Instead, focus on the URLs that get cited the most widely across different responses when you run a bunch of different prompts related to your product/service category.
Here’s why:
If you can get your brand mentioned in widely-cited, non-Wikipedia content, that content will probably continue to be widely-cited months from now.
Looking at that orange line, if you assume the same monthly decay rate of the most recent period (-12%), a widely-cited, non-Wikipedia URL will still be cited at more than 15% of its original frequency after a year.
That’s much better ROI than if you focus on the URLs cited by a single prompt.
As you can see in the graph, it’s an even bigger win if you can get your brand mentioned on a Wikipedia page that’s widely cited in your category, since it’ll likely be cited more widely in the future.
So how, exactly, do you generate “a bunch of different prompts related to your product/service category” in the right way to get widely-cited URLs?
How to find the widely-cited URLs in your category
I like to generate at least 50 commercial prompts for this exercise.
“Commercial” means a prompt should reflect someone asking an AI for product/service recommendations in your category.
The simplest and most generic version of a commercial prompt is, “What’s the best [product/service]?” For example, “What’s the best CRM software?” Or “What’s the best SEO agency in Chicago?”
These are the prompts for which AI models will consistently generate responses that mention brands.
These types of prompts reflect high-intent buying behavior. The responses to these prompts are the most important ones for your brand to appear in when you want to drive revenue.
But you can’t just stretch “What’s the best CRM software” into 50 different versions using synonyms and different sentence structures.
So you can use AI to create a list of 50+ commercial prompts, which is a process I’ll detail in a future newsletter.
Or, for ChatGPT citations, you can simply do a quick search in Contender. You can see the most-cited URLs in any B2B category in under a minute.
1. Sign up for Contender. It’s free.
Contender has a database of >100K commercial B2B prompts, with ChatGPT response data updated monthly.
2. Search for a product or service category.
Contender has ChatGPT response data for >2,100 B2B categories.
3. Click the “Sources” tab.
Voila. These are the most-commonly cited URLs in your category based on ~50 commercial prompts and recent ChatGPT responses.
4. Bonus: Get ideas for commercial prompts to track in your category.
Just click over to the Responses tab to get 40-50 commercial prompt ideas for any B2B category.
In the next issue, I’ll share my process for generating big lists of commercial prompts that cover all the scenarios your potential buyers might describe.
Cheers,
Mike
P.S. Need help getting your brand mentioned in the content cited by AIs? Reply to this email to learn more about our brand-mention-building service.