How to better measure LLM visibility and its impact

One of the most energetic conversations around AI has been what I’ll call “AI hype meets AI reality.” Tools such as Semush One and its Enterprise AIO tool came onto the market and offered something we could not live without: The data on what was happening inside LLMs. How many citations were we capturing? How many mentions were our brands gobbling up? 

This data became, and still is, an incredible novelty. But with the data came questions like “What’s the ROI here?” and “How do I interpret this data and integrate it into my team’s actual marketing strategy?” 

It’s clear the data provided by these tools is very valuable and very intriguing, but what do you do with it? 

The fundamental problem with tracking LLMs 

Why should tracking LLMs be any more problematic than tracking traditional metrics like Google rankings? The truth is that a KPI like Google ranking does suffer from similar shortcomings. Rank doesn’t equate to dollars. In fact, rank does not necessarily mean there is traffic coming to the site. If Google has an AI Overview right on top of your ranking, traffic may well be minimal for that keyword. And even so, is your site getting the right traffic to meet your business goals? 

The difference between traditional Google rankings and LLM visibility is that the equation demonstrating the relationship between strong Google rankings and increased revenue is not complicated. People search for a keyword, your site ranks at the very top of the results, and then users are more likely to click on your organic listing rather than someone else’s. 

From there, you can track how the consumer behaved. Once they landed on your site, did they leave in three seconds flat? Abandon their cart? Make multiple purchases after multiple visits?

Whatever the case, the connection is clear: SEO did its job by bringing people to the site. (Whether they were the right people is a conversation for another time.) Once we have people coming to the site from Google, we can track how good the CRO is or determine what needs to be adjusted in order to take the consumer from site visitor to site consumer. 

The transaction journey with an LLM is far more ethereal and disjointed. SEO and CRO work like brother and sister to create a transparent data journey. Conversely, LLMs are like your messy cousin who can’t figure out how to clean up after themselves at your grandmother’s 90th birthday party. 

Fundamentally, this is because Google Search was built to drive consumers to your ecosystem (i.e., your website). LLMs couldn’t give a hoot about your website. They are not built to deliver traffic as part of the pathway. Moreover, more and more data indicate what I feel has been obvious: LLMs are very top of the funnel. They are a starting point; a jumping-off point for users to further explore a topic or product. Evidence of this is found in basically every study under the sun, indicating that users tend to verify the information LLMs offer them. 

This construct makes turning LLM visibility data into something meaningful far more complex and layered than we’ve been used to with SEO data.

Simply put, because LLMs don’t drive users to websites (an ecosystem we own and can measure), tracking their role—and therefore their value—within the conversion funnel gets incredibly tricky. 

Great, my brand saw a 1,000% increase in the number of mentions it ranks for inside ChatGPT. What now? What does that mean? How can I qualify it? Is there a simple CTR-type of metric that says, “If you are the first brand mentioned in an LLM’s output, there is an 11.5% chance the user will visit your website?” 

No. 

So, now what? 

Now we get to where this gets really interesting. 

The problem with the methodology

Before we get into the weeds, I want to outline the basic methodology of how I approach utilizing LLM-related data. The issue with it is not how complex or simplistic the methodology for qualifying LLM visibility is or isn’t. It’s about how comfortable we are with it. 

The question of tracking LLM visibility in a more substantive way is not new. Not at all. Brand marketers have faced this problem since the dawn of the billboard. 

What makes this data process difficult for us is that we, as performance marketers, need to step out of our comfort zones for a moment. We’re very much used to direct attribution, direct data points, and not relying on correlational data. 

(I would venture that we’ve already been relying on correlational data, but this isn’t the place.) 

Regardless, the methodology around qualifying LLM performance is far more holistic, far more implicative, and far more correlative. Whereas performance marketers have historically been trying to create a data equation, they are now being asked to create an equation that fits into a certain picture or narrative rather than direct causation (or the appearance of it). 

That is new. 

Performance marketers are now being asked to create a narrative and a picture that shows the value of tracking LLM visibility. That picture is a little messier than the typical SEO or PPC reporting you may have done previously. 

However, marketers on the performance side are sitting on a wealth of information that their wider teams may not even be aware of. What you, as SEOs (and beyond), bring to the LLM visibility table is invaluable—once you get a little creative with it. 

The metrics and approach to measuring the impact of LLM visibility

Since the metrics included in LLM visibility reports don’t create the direct relationship and pattern performance marketers are accustomed to, discovering the value brought to your organization is a more layered process.

Before we dive in, it’s very important to understand what is happening within the full ecosystem of your company’s promotional efforts. If we’re creating a lot of correlated, but still deep, pictures and narratives, we have to know what should and shouldn’t be included in the data story, and how what’s included is best qualified. 

If, for example, your brand ran a huge TV ad campaign at the same time that your team “optimized” for LLM mentions, how do you identify the root cause of site traffic or branded searches on Google and the like? 

While it’s not impossible to identify causality, that can only happen when you have full awareness of the other activities your brand runs that may inform the same metrics you use to qualify LLM visibility. 

That caveat aside, what I tend to find is that the data from LLM visibility tracking is often the starting point for exploration and insight. This is unlike traditional SEO data, where the insights and end validation are much more closely seen, outward, and to a degree, “apparent.”

I’ll give you an example. According to the Semrush AIO Enterprise tool, TCL’s 6-Series of TVs is a trending product: 

A good 8% of the prompts I am tracking in the project pull up a brand mention related to the product line. 

With traditional SEO data, there was a tendency to stop here (not that this inclination was correct either, in my honest opinion). 

However, while ChatGPT seems “eager” to mention the product line, actual people are looking for it with less frequency. 

In November 2024, the keyword “tcl 6 series” had a global search volume of 6K and a monthly search demand of 3.6K in the US:

Those numbers fall considerably just one year later:

In November of 2025, global search volume for the product fell by over 35% to 3.8K monthly searches, while demand in the US fell to 1.6K monthly Google searches.

You can qualify this pattern in the branded traffic TCL receives from Google: 

In the above, the metric has essentially flattened. Going back to 2024, the brand has a consistent level of branded traffic from Google. 

Not so when you filter to only show TCL’s branded traffic as related to its 6 series TV product: 

In the above, you can see a serious decline in branded traffic associated with the product line, beginning in April 2024. 

At this point, we’re faced with a fundamental question: Does the product line have less interest because of a product issue, or is it simply getting outdone by a competitor? 

Qualifying this should be relatively easy. If I run a search for “is tcl 6 series a good tv,” the search results seem positive: 

While not perfect, the SERP points to the TV being a solid overall option. Google’s AI Overview confirms:

Since it’s not a product problem, it most likely means it’s a market/competitor issue. 

When we zoom out and look at how TCL stacks up in terms of LLM visibility across all of its TV-related products, you can see it lags behind powerhouses Samsung and LG by a good 10 percentage points, if not more: 

Now we have a starting point to run with. The next step would be to more fully audit Samsung and LG from their campaigns, activities, market perception, and the like in order to see where TCL could make up some ground with a product people really seem to like. 

My point is that using LLM visibility is far, far more layered and strategic than most of the traditional SEO metrics we’ve worked with for years. There are insights there, but you have to delve further into them and use them as red flags that indicate starting points where further investigation is required. 

If you can do that, you can set yourself, your team, and your company apart with the information gained. 

The branded search of it all

I want to focus more on using brand search to qualify LLM visibility. I touched on it in the section above, but I think brand search has always been a marketing insight goldmine, and now it’s even more valuable than before!

One of my favorite cases highlighting the value of how branded search data qualifies LLM performance is the example of the chicken wing chains Buffalo Wild Wings and Wingstop. 

The legacy brand Buffalo Wild Wings, which offers a dine-in experience (as opposed to Wingstop), has 18.4K mentions and 3.2K cited pages across LLMs:

Its newer (and takeout only) rival Wingstop (which has seen tremendous growth) pulls in about 4K fewer mentions but roughly the same amount of citations:

The question is, why? Well, Wingstop has about 200K more followers on both TikTok and Instagram: 

It is seemingly the cooler, hipper, more trending brand. Yet, its LLM performance lags behind its legacy competitor. 

To me, the branded search data tells this story. It’s not a matter of who is more popular for LLMs, but rather, who is more well-known across the internet (and all of the various concepts subsumed under that notion). 

Both brands have similar branded search traffic trajectories. Below, Buffalo Wild Wings draws in 5M visitors from Google for branded keywords:  

Almost exactly how Wingstop performs here: 

What’s different, however, is not the raw branded traffic numbers. 

This goes back to what I was saying earlier about using search data creatively and holistically to pull back the layers and create a narrative. 

Most of us would have focused solely on the number of visitors coming to each brand via Google through branded keywords. 

However, if we’re trying to get a complete picture of what LLM visibility data is telling us, we need to start thinking about this data more broadly.

In this case, I want to focus on the number of ranking branded keywords—not the traffic. 

In terms of raw branded traffic, both restaurant chains pull in around 5.8M users per month via some form or variation of their brand name. 

However, Buffalo Wild Wings uses roughly 360M keywords to get there:

In comparison, Wingstop gets its branded traffic numbers from Google with just 169K branded keywords. 

From a pure performance point of view, is this significant? You could argue it’s not. However, from a perception point of view and from an understanding of the awareness of these brands (which ties right into LLM visibility)—this is huge. 

Imagine if Nike got 10M visitors per month via branded search, but only one keyword was involved: “nike shoes.” That’s a lot of missed awareness. Nike also offers exercise clothes, sports jerseys, and a variety of products. That’s a lot of missed awareness that LLMs are going to pick up on. 

If you’re trying to use branded search data as part of your process to qualify LLM visibility, you should not only look at raw traffic but also qualify the keywords driving that traffic. 

What I see here with the two chains is very much a case in point. 

Let’s search for only branded queries that are related to “sauce” (as in “wingstop best sauce”). Wingstop has 406 keywords that yield 10.4K monthly visitors from the SERPs. 

Conversely, Buffalo Wild Wings’ branded sauce keywords amount to 3.7K for a SERP traffic yield of 70.5K: 

Which brand, from this point of view, is doing a better job with product awareness and diversity? 

Is it any surprise then that Buffalo Wild Wings has a larger total monthly audience in LLMs than Wingstop (with “monthly audience” being defined by Semrush as “total search volume for all topics where your brand is mentioned [in LLMs])?

Wingstop has a total monthly audience in LLMS of 56.8M:

That’s impressive, but it’s not nearly as high as Buffalo Wild Wings’ 98.7M:

And to me, that makes a lot of sense. Why? Because we took the initial LLM metrics presented to us and qualified them, then created a narrative that told the brand’s and the data’s story. So, coming back to a metric I purposefully glossed over earlier and seeing a discrepancy, it now makes total sense. 

This is how we need to approach measuring LLM visibility. It is the starting point. The data flagged that there may be an issue or an opportunity, and additional data was then required to contextualize and help us better understand what that issue or opportunity is. 

Direct traffic is your LLM data friend 

Branded search traffic and keyword search volumes are not the only metrics we can use to help us develop a fuller LLM data picture. It’s not even scratching the surface. There are all sorts of ways to give your LLM visibility data the narrative it needs. It can be anything from seeing correlations in how you resonate with your audience and who engages with your brand on social media to simply connecting the dots in your Google Analytics. 

One dot you could use in GA4 to help make that connection is your direct traffic. SEOs have long considered direct traffic a black hole, as anything that could not be properly attributed became “direct traffic.” 

While there is some truth to that, to me, direct traffic is fundamentally branded traffic. It indicates how well-known and adopted the brand has become that people simply type its domain into the browser and bypass everything algorithm-dependent (social, search, etc.). 

Tracking increases in LLM visibility alongside direct traffic can be a fabulous way to correlate the impact of AI brand awareness with actual consumer impact. 

Let’s go back to the earlier example we used with TCL and LG.

Again, LG performs on a very different trend line than TCL does: 

And that makes good sense to a degree, because LG is simply a bigger brand. 

You can see this play out in the direct traffic comparisons between the two: 

LG (green line) is pulling in over 20M potential consumers per month via direct traffic. That’s mammoth compared to TCL. 

We can also see that LG is starting to pick up the amount of traffic it earns from LLMs:

As this happens, and as we see LG gaining more momentum in LLMs, I would pay attention to the direct traffic metrics. If the trends of brand mentions and direct traffic start to move together in the same direction and at similar rates, that could be a positive signal that the LLM visibility is having a practical impact. (Again, you may not see the impact of LLM visibility in exactly this way, as the impact might instead appear in other metrics like increased branded search or others.)

However, I would not rely on just one metric. Again, we’re creating a data picture. In this scenario, we’d have to determine if other factors (perhaps paid media) are increasing and driving direct traffic. Also, looking at other metrics like overall demand (via keyword search volumes) and branded traffic should all come together to paint that picture for you.

It’s not about only one metric

All I’ve done here is show you how I go about investigating and creating a narrative from LLM visibility data. This narrative is essential as it provides more depth and can be better attached to specific goals and actual KPIs. 

In the case illustrated here, I used search volume data, branded search data (in a very non-linear way), and some direct traffic data. But really, anything and everything can help tell your LLM-visibility-data story. Are bounce rates down in accordance with more LLM visibility? Does that help you gauge whether your audience targeting via LLM output is better than other places across the web? 

Whatever it may be, the idea is that there is a wealth of performance-focused search and marketing data available that you are already an expert in. You can bring that to the wider team and drive value with it. You have more insights available to you than mentions and citations.

While those metrics are a great start and a huge piece of the puzzle, it’s what you do with that data and how you supplement it with the rest of your performance data stack that will provide the opportunity to showcase your expertise and added value.