Here's a conversation I seem to have more and more…
Marketer: "Can you help us win in LLM search?"
Me: "Sure! How is your LLM search performance at the moment?"
Marketer: “We were hoping you could tell us.”
So that’s what this blog post is about. I’ll show you the metrics that I use to measure LLM visibility and how I do it.
LLM leads are pretty high-intent
Apart from the AI hype, there are plenty of reasons to start competing in LLM search.
In terms of sheer traffic volume, Google is still generating a lot more than LLMs. But the conversion rate from LLM traffic converts at higher rates – for some of our clients, that conversion rate from LLM search is 10x compared to Google.
So we want every client’s brand to be visible in that channel. We want to provide helpful answers to the questions that people are asking in ChatGPT or Perplexity or Gemini. When those LLMs talk about your brand, we want the answers to be accurate. And we also want people to click through to our site from the link citations.
Overview: what exactly is “LLM visibility”?
How frequently a brand is mentioned in LLM search results about certain topics.
👆That’s how I’m defining it.
We measure LLM visibility with a percentage. For example: Our brand is mentioned 45% of the time when people ask questions about document processing.
The goal is not necessarily to be mentioned 100% of the time. The goal is to be mentioned more often than your competitors.
Note: LLMs are not deterministic. The answers they give to questions change slightly every time. And so this visibility percentage is going to fluctuate from month to month. (Not so different from Google, in that regard.)
If you’d like to get in the weeds with me here, check out the full blog post I wrote. Also, you can follow along with this video 👇
How I track LLM visibility
Basically, we’re going to run a bunch of targeted queries through some popular LLMs and quantify how frequently certain brands are mentioned. We’ll combine and average out the mentions from the various LLMs to get a visibility metric (a percentage) for each topic.
Here’s the basic process:
- Pick a list of strategic topics. These are topics that align with product marketing and audience pain points.
- Identify a variety of queries within each topic. Think of these as subtopics, search queries, or (to use some SEO lingo) long-tail keywords.
- Run those queries through major LLM tools. My workflow runs them through ChatGPT, Gemini, and Perplexity.
- Document and quantify the responses. We’ll quantify the number of queries in which our brand appears in LLM responses. We’ll also quantify the mentions of top competitors. And we’ll document the responses to run quality-assurance tests on the data itself.
We use a straightforward process to gather our LLM visibility data. In essence, we’re going to create a batch of targeted queries then run each one through popular LLMs and quantify how frequently certain brands are mentioned.
We’ll combine and average out the mentions from the various LLMs to get a visibility metric (a percentage) for each topic.
Here’s the basic process:
- Pick a list of strategic topics. These are topics that align with product marketing and audience pain points.
- Identify a variety of queries within each topic. Think of these as subtopics, search queries, or (to use some SEO lingo) long-tail keywords.
- Run those queries through major LLM tools. My workflow runs them through ChatGPT, Gemini, and Perplexity.
- Document and quantify the responses. We’ll quantify the number of queries in which our brand appears in LLM responses. We’ll also quantify the mentions of top competitors. And we’ll document the responses to run quality-assurance tests on the data itself.
We run these tests with our own LLM workflow, as you may have guessed.
However, you can get a feel for the process by a little bit of manual research for yourself.
Open up one of those LLM tools. Type in some phrases that are strategically important to your company’s marketing. See what (and who) comes up. It can be very enlightening.
The output of my report: visibility metrics
At ércule, we help companies dial into the handful of topics they really want to own. This lens has always been super useful in organic search. Now we’re seeing the importance of topic visibility in LLMs skyrocket. When we start to set up a topic strategy with a client, we want to give strong consideration to both angles.
For this example, we’re working with a company that is selling an AI-powered documentation tool. Here’s a spreadsheet that lays out the data for ten strategic topics.
In this particular case, when LLM users ask questions about “agentic document processing,” this brand is mentioned 33% of the time. For context engineering, it's 35% of the time.
To frame the insight from this metric a little differently: if I ask about context engineering in an LLM then this particular brand is going to show up approximately one out of three times.
What kind of data are we basing this on?
When I click on my little “Data” tab, it shows me the documented queries and responses collected by the query workflow.
Within each of the topics in the left-most column, our workflow has sent a variety of subtopics or longtail queries to the various LLMs. Within the topic of context engineering, we searched this phrase: “best content engineering tools.”
Our workflow plugged that into GPT, Gemini, and Perplexity. We collected the entirety of each LLM’s narrative response. This tab provides quick previews of the data. Responses that begin,
“The best context engineering tools in 2025 focus on optimizing how AI systems…” etc.
We keep query records for quality-assurance
Having this data available at a glance is mostly useful for quality-assurance purposes. We want to make sure that the prompts we’re testing are actually relevant to our brand.
For example: within the “content engineering” queries, our workflow tested a query about social engineering.
What is social engineering within the context of information security? Not a relevant prompt for our company. So this instance needs to be struck from the data set. Otherwise, it’s skewing the overall visibility metrics.
That’s why we catalog all the responses.
Compare your visibility against competitors
The other part of the equation is: how are your competitors doing?
Once you’ve gotten responses from LLMs, you can use the same data to check competitor visibility. In our LLM visibility spreadsheet we do a simple check: “Does the LLM response contain the brand name?”
The metric here is the same as the original LLM visibility: what percentage of the responses include mention of this brand?
In this example, you can see that our brand is appearing more often for some topics rather than others.
Refining the report
We can get more granular with this kind of report, too. We can scan for things like…
- Citations. Which company websites are being cited most often in LLM responses? This has big implications for inbound leads.
- Sentiment analysis. Are the mentions saying good things? Are they saying bad things?
- Brand alignment. Does the LLM’s description of your product match your own preferred messaging and positioning? Does it quote outdated marketing pitches?
And when you start paying real money for software, you can also refine everything further. You might qualify the search responses for different buyer personas, for example.
This is all part of one unified content system
For all the talk of the great AI takeover, Google search data is just as important as ever.
LLM search and Google search are two channels within one cohesive content system. They’re two entry points to your website.
LLM is still a growing channel. The volume isn’t huge but the leads that it generates are significant.
The goal is to be accessible to your audience on every viable channel, be it LLMs or Tik Tok or good old email. The nice thing about LLM search is that you don’t have to adapt your content to a radically new format. You just have to optimize what you already have.
Next steps
Before you pay for a fancy LLM search tool, type a few strategic queries into GPT or Gemini. See what's happening for your key topics. Ask them what they think of your company and your competitors.
If you’re seeing your competitors much more often than you’re seeing your own brand mentioned, document those results. If you spot a trend, start tracking it in a spreadsheet. A simple system you use is far better than a complex system that you never touch. And if you’re interested in going the DIY automated route, send me a DM anytime.
