You run a tool that measures AI visibility. The result: you're barely there. Then you open ChatGPT yourself, ask “which [your category] should I choose” — and it recommends you. Or the reverse: the tool shows a nice score, your domain keeps landing in the sources, and yet your brand isn't in the recommendation itself. In both cases the same thing is happening: the tool measures one thing, and you think it measures another.
Domain visibility and brand visibility are not the same thing. Most tools — especially the self-built ones, and there are more of those every month — measure only the first. And that's a problem, because for most brands the second one matters more.
Two things we keep confusing
Before we measure anything, you have to split apart two questions that blur into one in your head.
First: does AI cite your domain? That is — when a model builds an answer, does it reach for your URL as a source. That's domain visibility. Measurable, concrete, and it's what the DIY tools catch: they check whether your address shows up in the answer's sources.
Second: does AI recommend your brand? That is — does your brand appear in the body of the answer as a recommendation, regardless of where the model got that information. And very often it gets it from someone other than you. From a comparison site, from a ranking in an independent article, from a review, from a comment. Your domain isn't in the sources at all — and the brand is recommended.
Definitions · remember this difference
Domain visibility — how often your URL is a source AI reaches for when it builds an answer.
Brand visibility — how often your brand is named and recommended in the answer, regardless of which source the model took that from.
The distinction sounds academic until you see what it does to your numbers. And it does plenty — because these are two completely different contests. That's where the real story starts.
These are two separate contests
This is where the whole difference sits, and why a single “visibility percentage” from a tool can mislead you.
When you measure domain visibility, your domain competes with every domain on the topic. Not just competitor sites. You're up against publishers, portals, blogs, comparison sites, rankings, forums, editorial desks. When a model builds an answer, it picks sources from the whole internet on that topic — and in most categories it's the independent publishers and aggregators that have the thicker, older, more “citable” content, not a single brand's site. It's a crowded, hard arena. Your domain may barely register in it.
When you measure brand visibility, your brand competes only with other brands in the category. That's a narrower field. You're not racing Wikipedia or a “top 10” ranking — you're racing three, five, ten competitors. And suddenly it turns out you're recommended more often than your domain's presence in the sources would suggest.
That's why the brand score is usually higher than the domain score. It's not a measurement error — it's two different contests. The domain fights everyone. The brand fights only its own shelf.
The practical takeaway: if you only look at whether your URL is a source, you understate your real picture in AI. Because your brand may be recommended over and over — purely on the strength of what others wrote about you.
Why DIY tools measure the wrong thing
There are more AI-visibility tools every month, and a good share are self-built — a script that queries the model with a set of prompts and checks whether your domain shows up in the answer (or its sources). There's one fundamental shortcut in that: such a tool only sees the domain as a source. And that creates two problems.
Problem one: a bias toward informational and educational prompts. A brand's domain lands in the sources most easily on questions like “what is X” or “how does Y work” — because that's where you have a blog article a model can quote. So the tool measures you best exactly where it matters least commercially. The prompt “which [product] should I buy” or “recommend a [service] for [situation]” — the one you actually make money on — very often doesn't cite your domain at all. It cites a ranking, a comparison site, a review. And the DIY tool shows: zero. Even though your brand might be the first recommendation in that answer.
Problem two: it misses recommendations from third-party sources. This is the flip side of the first. The brand is recommended, but the source is an independent domain — a comparison site, a ranking, an article, a comment, a review. Your site is nowhere near it. A tool that looks only at the domain records this as “you're not visible.” In reality you are — just not thanks to your own site.
AI measurement tool · weak vs good
Weak tool
- Only counts whether your domain is a source
- One “visibility” number
- Confuses no-domain with no-visibility
- Skews toward informational prompts
Good tool
- Measures brand visibility first, regardless of source
- Shows separately whether the domain is a source
- Analyses brand perception — how the model describes you
- Separates prompt intents
Which leads to a simple rule I keep repeating: if a tool can't catch a brand recommendation with no link to a domain, and doesn't analyse brand perception, it's not a good tool for measuring AI results. It measures one layer out of three and reports it as the whole.
Why legacy brands win in AI, even though they “can't do SEO”
Now the most important part, because it explains why any of this is worth separating.
Look at who AI recommends in your category. Very often they're not the brands best at SEO, or even at online marketing. They're the brands that are simply known. They built recognition back when the TV ad was what mattered — sometimes earlier — and ever since, a lot just gets written about them. They don't have a brilliant site, they don't have a slick content strategy. And they win anyway, because the model understands one thing: to the user, this brand is the default choice.
This is the thesis I repeat on this blog constantly: the brand is the substrate of visibility in AI. Optimisation — SEO, GEO, AEO — works on recognition that already exists. It can't create it from zero. The model reconstructs what the world already “knows” about you: how much is said about you, in what context, how consistently.
And here's the caveat, so you don't draw the wrong conclusion. This does not mean “do nothing and you'll be visible anyway.” It means the opposite. Recognition can't be faked with on-page tweaks — but it can be built, and that's the controllable lever. The lever is how much and how you're written about beyond your own domain: Digital PR, presence in rankings and comparison sites, mentions, reviews. Legacy brands built that over decades. You have to make it up deliberately.
So optimising your own site is one thing, but Digital PR is just as important — and in many categories more so. In AI, the winner is often not whoever was best at SEO. It's whoever the world has the most consistent things to say about. That's work no single SEO specialist delivers alone from a chair next to the site — it's a team, a budget, and a long horizon. But it's what decides.
What it looks like in a tool that does it right
I measure this in chatbeat.com, and that's exactly where I have it split into layers — not one number, but separate questions treated separately.
In the first layer, visibility is scored on the brand, not the domain. That is: how often and how strongly your brand is named and recommended in answers — regardless of where the model took that from. Here your brand competes with competitor brands.
In the second layer, it separately scores whether your domain is a source of the answer. That's a different league: here the domain competes with every domain in the topic space — publishers, comparison sites, rankings. That's why the domain score is usually lower than the brand score, and that's normal.
Two layers of measurement
| BRAND layer | DOMAIN layer | |
|---|---|---|
| What it measures | Whether the brand is recommended | Whether your URL is a source |
| What it competes with | Competitor brands | Every domain on the topic |
| Question it answers | “Does AI recommend me?” | “Does AI cite my site?” |
| Why look | Commercial picture | Narrative control |
On top of that comes a third thing a good tool shows alongside visibility itself: brand perception — not just whether you're recommended, but how the model describes you. That's a separate topic I break down in the piece on brand perception in AI answers — here it's enough to remember that visibility without perception is still only half the picture.
What to do about it — and when the distinction doesn't matter
Depending on how your brand stacks up against your domain, the conclusion is different.
Strong brand, weak domain
You're recommended, but mostly by other people's hands — from rankings, comparison sites, reviews. You have the substrate; what you're missing is control over how you're described, and reach where third-party sources don't go. Your domain is your biggest untapped opportunity: your own content that the model can cite directly. That one's on-site work — and it's where making your pages citable pays off most.
Weak brand, weak domain
There's no shortcut here. On-site optimisation helps at the margin, but the real work is building recognition: Digital PR, presence, being written about. It's a long game — a return to Inbound Marketing fundamentals — not a one-quarter campaign.
When you don't need to agonise over the distinction
If you sell locally, in a narrow niche where only a handful of specific queries matter anyway, a simpler measurement is enough. Separating brand from domain starts to matter once you're genuinely competing for a recommendation in a category with enough to write about. If nobody in your industry is recommended by AI yet, you don't need a sophisticated tool for it — you need to start existing.
One thing is certain: as long as you look only at whether your domain is a source, you're grading yourself in AI on the least important of three layers. Start with “is my brand recommended,” and only then “is my site the source.” In that order.
FAQ
Frequently asked questions
It matters — it's just not the same thing, and not always the most important one. Domain visibility gives you something a recommendation from someone else's source can't: control over how you're described, and a direct citation where third-party content doesn't reach. Treat it as the second layer, not the whole score.
Because it takes the information from elsewhere. Models build an answer from the whole corpus they have about your category — rankings, comparison sites, articles, reviews, mentions. If the world writes a lot and consistently about you, your brand shows up in the recommendation even when your domain isn't in the sources.
Because they're two different contests. Your brand competes only with competitor brands — a narrower field. Your domain competes with every domain on the topic, publishers and comparison sites included — a much wider field. A brand score higher than the domain score is the normal, healthy picture. The reverse — strong domain, weak brand — usually means you're cited but not recommended, which is a signal to work on recognition.
Not useless, but incomplete. If it only measures whether your domain is a source, you get one layer out of three — usually the one least tied to revenue. Before you draw conclusions from it, check whether it can catch a brand recommendation with no link to your domain. If it can't, don't judge your real visibility by it.
Yes. Your site is the one layer you fully control — and the only place a model quotes your words rather than someone else's. Optimisation won't build recognition from zero, but it converts your existing brand potential into direct citations and narrative control. It's complementary to Digital PR, not a substitute for it.
With existing, not with a sophisticated measurement. If nobody in your category is recommended by AI yet, you don't need an advanced tool — you need the world to start writing about you: Digital PR, presence in rankings and comparison sites, a consistent narrative. The brand-vs-domain measurement starts paying off once you're actually competing for a recommendation.
