Does the person who looks after your website ever talk to you about how your website is showing up on AI searches? There’s a serious risk that the measure you are being shown is not the one that’s most important.
The tools that do the tracking are called “citation tracking tools”, and they measure whether or not AI produces a link to your website. That’s good. Isn’t it?
Think of the end user getting these links. They click on them, have a quick look, and move on to the next one. But they won’t remember a domain URL so they can accurately go back to it and follow through reliably. People do that by remembering brand names — that is, by remembering the name of your business. That’s what matters commercially. The number of citations (domain links) you get is not the key point. The number of brand mentions you get is. And the measure of citations can be significantly misleading as to how well you are doing.
This mismatch is supported by independent research. Kevin Indig’s study of almost 4,000 domain appearances across four AI engines found that three-quarters of the domains received citations — source links showing in the panel — but fewer than 40% received brand-name mentions. And most importantly of all, fewer than one in seven actually had both.
This is supported by further evidence from Semrush, which shows that fewer than one in five brands are both mentioned and cited. This isn’t accidental, and it isn’t a passing phase. It’s structural — it’s built into the way AI works.
We have spent a year looking into this and trying to understand exactly how it works. Most people assume that if your domain name turns up in a response to a user’s question, then the brand name follows. As you’ve just seen from the evidence, that is not the case.
The opposite is more true, however. If your brand is mentioned in a response, then the chance of the domain name turning up as well can be better than 50%. If your brand isn’t mentioned, that chance falls to around one in ten.
The reason is rooted in the way AI works. The AI doesn’t start from your content and work outward. It starts from its training knowledge — a picture of which brands are associated with which categories, built from the whole of the web before the model was switched on. That “training knowledge” layer decides which brands to recommend. A second step then goes looking for source material to cite alongside the answer. If your brand name was in that source material, your URL surfaces as relevant.
An important caveat: this is a general description of how AI works. Each engine has its own way of handling it, and each also has access to up-to-date information that isn’t yet in its training knowledge. That’s why, when you run the same query several times — one after another, or on different days — you’ll likely get different answers.
If your brand isn’t recognised in that training knowledge as a relevant player in your category, chances are that citations alone will not produce brand-name appearances. The AI may well cite your work as an anonymous source while recommending your competitors by name.
Your content is working. Your brand isn’t. You are involved in two different battles here, and most people only see one.
The two battles, in plain terms
The first battle is your Reputation — what the web says about you. It’s built from everywhere your name comes up: reviews, forums, social posts, industry write-ups, the general chatter of the web. It isn’t built from your website alone. This is the evidence AI-powered search engines draw on, and the more factual, consistent, and well-organised it is, the more weight it carries.
The second battle is your AI Visibility — whether you actually get named. When a customer asks an AI-powered search engine who the best businesses in your field are, either your name appears in the answer or it doesn’t. This is the battle that decides whether you get recommended at all.
We measure the two separately because they move independently. A business can have a solid reputation and still go unnamed, or get named on thin evidence that won’t hold. Only measuring both tells you which one is holding you back.
Here’s the catch. The usual tools measure the links — the citations — so that’s the number most businesses end up chasing. They pour effort into content, watch their citation count tick up, and never notice that their name still isn’t appearing in the answers customers read. Fewer than one in five brands win both battles at once — and that overlap is where all the commercial value sits.
Why one “AI visibility score” is misleading
There is no single AI visibility number, because the different engines behave almost like opposite personalities.
Some engines link to sources constantly but name hardly any brands — think of the endless academic footnote, plenty of references, no names. Others do the reverse: they’ll happily recommend brands by name and rarely bother attaching a formal link. In Indig’s data these differences were stark — one major engine named brands roughly four times as often as another for the very same searches.
So a report that says “your AI visibility is healthy” based on citations from one engine can be telling you almost nothing about what’s happening on the others. In a fifth of cases across engines, the AI systems didn’t even agree with each other on whether to name the same brand. One engine would recommend a company by name every time; another would quietly use it as an unnamed source every time. Same business, completely different treatment. That’s four separate scoreboards, not one — and a single headline figure hides it.
The kind of content you publish decides which outcome you get
This is the part most businesses get wrong without realising.
Explainer and reference content — the “how does this work” material — tends to get used as an anonymous source. The AI lifts the information, answers the question, and never names you. Useful to the AI; invisible to the customer. In the research, some big reference sites were cited dozens of times and named almost never.
Content that makes a judgement — named comparisons, recommendations, “who’s best for this” pieces, case studies where your name is attached to the result — is what actually produces brand-name mentions. When your name is inseparable from the point being made, the AI can’t use the point without saying who you are.
The type of question matters too. The same business can show up in half of all “what’s the best…” answers and almost never in the “how does… work” answers on the exact same subject. So if the tracking you’re shown leans towards the explainer-style questions, you’re being measured on the part of AI search that matters least commercially — and missing the part where recommendations are actually made.
What you should actually be watching
The single most useful thing to measure isn’t your citation count. It’s the gap between how often AI uses your content and how often it actually names you. A large gap — lots of links, few mentions — is the warning sign. It means your material is good enough to borrow but your brand isn’t strong enough to recommend.
One honest word of caution: AI doesn’t change its mind overnight. When you improve your content and your wider presence, it typically takes something like four to eight weeks for that to feed through into the answers. So this is a trend to watch over months, not a dial that moves the day after you make a change. Set your baseline now, and track the direction of travel.
What actually closes the gap
There are three things that move the needle, and they work together.
Put your name inside your best claims. This works on your AI Visibility. The most immediate lever, and one you control entirely. Write your content so the insight and your business name can’t be separated — if the AI can’t make the point without naming you, it won’t. This is editorial discipline, not new infrastructure, and it’s the surest way to turn a use of your content into your name actually appearing.
Build your presence as a recognised name in your field. This works on your Reputation. The slow, foundational work — being consistently and clearly identified across the web as who you are and what you do, so the evidence AI draws on all points the same way. It’s what drives the whole “get talked about, then get recommended” effect. It takes longest and matters most.
Get named by others in the places AI trusts. This builds your Reputation and, in turn, your AI Visibility. Independent coverage, comparisons, recommendations and “best of” mentions from credible sources. The lion’s share of AI’s citations come from this kind of earned coverage, not from your own website — which is exactly why on-site content alone can never win it for you, and why a credible third party naming you is the strongest signal of the lot.
Do only the first, and you’re polishing the footnote. The other two build your Reputation — the evidence AI leans on — and that Reputation is what turns into the AI Visibility that actually brings customers in.
The bottom line
This is why we measure two separate things: your Reputation — what the web says about you, the evidence AI draws on — and your AI Visibility — whether you actually get named in the answer. The gap between the two is the whole game, and most of the tools and reporting you’ll be shown only ever measure one side of it. They tell you whether AI finds your content useful. They don’t tell you whether AI is actually sending customers your way.
If the report you’re looking at counts links but not mentions, it’s counting almost everything except the thing that matters. Fewer than one in seven AI appearances deliver both. Knowing which side of that line your business sits on is where real AI visibility work begins — and it’s exactly the gap we help businesses close.
Sources: Kevin Indig’s study of 3,981 domain appearances across four AI engines; Semrush’s analysis naming the “Mention-Source Divide”; Seer Interactive’s analysis of 541,213 AI responses; Ahrefs’ analysis of AI source behaviour.