How Demand Genius Tracks AI Visibility Across Your Entire Brand Footprint

Picture a Tuesday morning. You open your AI visibility dashboard and the line has dipped. Not a catastrophe, but enough that your CMO has spotted it first and dropped you a message asking what happened.

So you dig in, and what you get back is a sentiment score that shifted. No prompt, no source, no URL, no date. Nothing you can point to and say, “This is what changed, and here is what we are doing about it.”

That moment is why teams reach for a pure visibility tracker. The pitch works because it is clean. Your URL got dropped from a citation. Here is the exact data point. Go and fix it.

And somewhere along the way a trade-off got baked into how people compare tools in this category. Granular tracking data on one side, strategic context on the other.

We think that trade-off is false. At Demand Genius we track your visibility down to the individual prompt, the entity, and the exact URL an LLM cites, and then we show you why any of it moved. The clean data point and the bigger picture were never competing features. One sits on top of the other.

Let us walk you through what we actually track, how attribution works when your numbers move, and why a B2B brand needs more than a mention count.

Your Brand Is More Than Its Name

Before we get into what we track, we want to challenge how you define the thing being tracked in the first place.

Most visibility tracking starts from a single assumption. Your brand equals your brand name, so measurement means counting how often that exact string shows up in AI responses. That holds up for a consumer brand where the name carries everything. For a B2B brand, it misses most of what an LLM actually says about you.

Take us as an example. Demand Genius, the name, is one part of our footprint in AI. Dark AI, the concept we coined, is another. So is Tom, who authored the research behind it. So are canon concentration, our benchmark reports, and the category language we have pushed into the AEO conversation.

When an LLM explains the dark AI phase to a marketer asking why their organic traffic looks flat while pipeline holds steady, we are present in that conversation even if the answer skips our name entirely.

Now run the same exercise on your own brand. Your product names. Your founder or the executive whose LinkedIn posts get quoted back to you. The framework your team coined three years ago that competitors now use in their own decks. The benchmark report that analysts reference. Every one of those is a surface where AI talks about you, and a name-only tracker records silence across all of it.

This gap is wider than it feels. When we analysed hundreds of prompt clusters for our research on dark AI, we found that brand mentions in awareness and consideration prompts are far more fragmented than at the decision stage.

The model references pieces of your brand, a framework here, a person there, long before your name appears next to a recommendation. Tracking the name alone shows you just that final step.

When we say we track your visibility, we mean the whole footprint. Here is what that looks like in practice.

What Demand Genius Actually Tracks

Every AI answer your buyer reads gets assembled from parts you can measure. A question goes in, phrased one of thousands of possible ways, bundled with layers of context your buyer never sees. The model decides which names belong in the answer, with your brand among them or missing entirely. And it leans on specific sources to decide what to say about the names it picked.

Visibility lives in those three places, which is why we track all three. Prompts tell you where you show up. Entities tell you which parts of your brand do the showing up. Cited URLs tell you which pages earn your presence and which are losing their grip.

Visibility Trends Across the Questions Your Buyers Ask

There is a property of LLMs that shaped how we built this entire layer. Ask the same model the same question ten times and you will get ten differently worded answers, sometimes with different brands in them, because every output is generated fresh rather than retrieved from an index. A single prompt checked once a week is a coin flip dressed up as a metric.

That is why we run each prompt repeatedly across ChatGPT, Claude, Gemini, and Perplexity, and your visibility reads as a trend rather than a snapshot.

The clustering sits on top of that, and it is the part of our tracking we hold the strongest opinions about. We group prompts around the job your buyer is trying to get done, with up to 25 conversational variations per cluster, written the way real people ask.

We build them this way because a B2B journey holds two kinds of jobs. Product jobs, the things your tool accomplishes, and buyer jobs, the work of building a requirements list, understanding the category, building consensus across the buying committee, and making the business case for budget.

Tracking that only covers product jobs confines your measurement to bottom-of-funnel comparison queries, and the set of options a buyer considers gets decided well before those queries are typed.

Our clusters carry both, so the trend you read covers the whole conversation and one oddly worded prompt cannot swing your score.

It is also why we refuse the industry’s favourite shortcut, which is scraping your own site to generate the prompt list. A list built from your content is biased in your favour by construction. You end up measuring the questions you already answer and scoring well on your own homework.

We build the clusters from your buyer’s jobs instead, and your team reviews every prompt before anything runs.

The Entities That Make Up Your Brand

This is where the footprint from the last section becomes a tracked capability. In the platform, you pick out the specific entities that carry your brand beyond its name, the same surfaces you listed when you ran that exercise on your own brand, and we follow each one separately across every prompt and cluster.

Our dark AI research found that early-stage answers mention brands in fragments, a framework here, a person there, with the full name often arriving only near the decision. Entity tracking is how those fragments become signal.

When the model explains your framework to a buyer without printing your name, a name tracker records that nothing happened. An entity trend records exactly that event, on that day, in that cluster.

Separate trend lines also change what you can decide. Say your brand name holds flat across a cluster while your CEO climbs through awareness-stage questions. You now know the thought leadership is working and deserves more fuel.

Or your framework spreads through category conversations while your product entity stays still, which tells you the language won but the association back to you needs strengthening.

The Exact URLs AI Cites When Talking About You

Here is the data point pure trackers sell as their headline, so let us be precise about how it works in Demand Genius. For every tracked prompt, we show you which specific URLs drive the citations in the answer. We show you the same for sentiment, so you can see which sources the model leans on when it describes what you are and what you are good at.

When a page of yours drops out of a Perplexity citation, you get the page, the prompt, and the date. Nothing abstract about it.

Whether you can trust a data point like that comes down to how the visibility data gets made, and this is where trackers quietly diverge. Some query the model through a raw API, which strips out the product’s live retrieval and citation behaviour. Others puppet a consumer interface through managed browser accounts whose accumulated history belongs to no human on earth.

We collect through the API and then rebuild the context layer deliberately, configured to your ICP segments, so when a citation appears or disappears, you know the model’s behaviour changed rather than some anonymous account’s history drifting.

That standard is one we think you should hold every tracker’s numbers to. Including ours.

When Your Visibility Changes, You Can See Exactly Why

Go back to that Tuesday morning. The line has dipped, the CMO is waiting, and the difference between panic and a plan comes down to whether your tooling lets you follow the number all the way down to its cause.

We built every layer in the platform to be dug into, and that is a deliberate design position. Every trend line sits on top of the actual prompt responses, citations, and sources that produced it.

So here is what the dig looks like. Your visibility dropped in one cluster. You open it and see which prompt variations moved and which held steady. From there, a moved prompt shows you the responses the model actually gave, this week set against last. Then the citations tell you which URLs entered or left, and which sources the model leaned on when it framed its answer.

Three clicks in, “visibility dropped” has become something you can say out loud in a meeting. This page fell out of citations for these prompts on this date. Or a competitor’s new comparison guide started appearing in the sources for your cluster. Or nothing on your side changed at all, because the model itself moved.

That last cause deserves its own paragraph, because brand positions in AI drift even when you change nothing. Search Engine Land tracked project management software across September and October 2025 and watched Slack drop 8.1 points of AI visibility in a single month while Atlassian gained 5.5, with no obvious campaign behind either move.

When that happens to you, explainable tracking is what separates “the model shifted, here is our response” from an unfair post-mortem of your content team.

Sentiment gets the same treatment. When the way AI describes you changes, you see the sources driving the new description, side by side with the old ones. A sentiment shift you can trace to its sources is a to-do list. One you cannot is just weather.

Why Visibility Tracking Alone Isn’t Enough for B2B

Everything to this point has made one argument. We track visibility to the same granular standard as any dedicated tracker. Now we want to tell you why we refuse to stop there.

A citation report, even a perfectly explainable one, tells you where the model ended up. It cannot tell you whether anything you did put you there.

Here is what we mean. In our research, we ran the same buying questions through LLMs over and over, and measured how often the same brands came back each time. We scored this on a scale from 0 to 1, where 0 means the model named different brands on every run, and 1 means it named the same brands every single run.

On early, exploratory questions the score sat at 0.37. The model was still open, cycling different brands in and out of its answers. On decision-stage questions the score hit 0.82. By that point the model had settled on a small set of brands and repeated them almost every time.

Citations only start appearing at that second stage, after the model has already settled. A strong citation score usually tells you that you were already inside the model’s settled set of options. It says very little about whether your content earned that position or just inherited it.

And that settled set gets chosen where trackers are silent. Our research found that 84% of the prompts across a typical B2B buying journey produce no brand citation at all. Those are the awareness and consideration conversations where your buyer works out what the category is and which criteria matter.

A tool that reports on citations has nothing to report across that entire stretch, and that stretch is where the decision forms.

Then there is the question of who is asking. A B2B purchase runs through a committee, and AI works as a personalised adviser for each person on it. The CFO asking about auditability and the VP of Engineering asking about API flexibility get different answers, with your brand framed differently in each, or missing from one entirely. A single blended visibility number hides exactly which stakeholder you are invisible to, and that gap costs you deals.

None of this makes tracking optional. Clean, explainable visibility tracking is the foundation, which is why we built ours to the standard this piece has described. But visibility and influence are not the same thing, and the tracking layer only measures the first one.

So we measure what sits above it too. How AI frames your category. Whether it connects you to the criteria each stakeholder actually cares about. And whether your content is shaping the early conversations where the model settles on its set of options. On top of the tracking, never instead of it.

See Your Entire Brand Footprint in AI

Right now, an LLM somewhere is describing your category to a buyer you have never met. Some piece of your footprint is in that answer, or missing from it, framed by sources you have not read. That conversation happens whether you track it or not. The only thing you control is whether you can see it.

Being able to see it is something we have invested in heavily. We have just finished a whole load of work on the visibility layer of our platform, because we want it to stand next to any dedicated tracker on the market. Everything this piece has described is the result.

Book a call with us. It runs for 25 to 30 minutes, and we will pull up your brand live. You will see what is driving your visibility, what is limiting it, and you will leave with a clearer view of your AI positioning whether we end up working together or not.

And the next time the line dips on a Tuesday morning, you will have the answer ready before the message from your CMO lands.