Being Quoted Is Not the Same as Being Known
There are two kinds of AI visibility and they are easy to confuse. One is AI quoting your website. The other is AI knowing who you are. A brand can have plenty of the first and none of the second.
Key Takeaways
- Being quoted often and being known are different things, and one number hides the difference.
- Vendor A is the most quoted company in the study, yet only 5.6% of its mentions survive without its own website.
- Vendor C has 17 times fewer citations but 74.4% earned recognition.
- The difference is other people: 141 pages across 49 websites name Vendor C. Vendor B has none.
- Publishing wins the answers where your page is best. Being discussed wins all the others.
Why we looked into this
Most AI visibility reporting counts one thing: how often your website is quoted. It is easy to measure and it feels like the right number, because being quoted is visibly good.
But sit with an AI answer for a moment and you notice something that number cannot see. Sometimes the engine names a company because it is reading that company's website. Sometimes it names a company simply because it knows the company exists and belongs in the answer, with no link to them anywhere in sight.
Those are entirely different commercial positions, and one counter cannot tell them apart. We wanted to separate them, because we suspected a lot of brands are being told they are doing well on a measure that hides their actual weakness.
How we tested it
For every one of the 6,601 answers we recorded two separate things. Which companies the engine named in the text of its answer, and which websites it quoted as sources.
That lets us split every brand mention into two kinds. A mention that arrived alongside a citation to the brand's own website, where the engine was effectively reading their material and repeating it. And a mention with no citation to their own site at all, where the engine named them from what it knows about the category.
We call the second kind earned recognition. It is the share of a brand's mentions that survive when you take their own website out of the picture. We also counted how many other companies' pages mention each brand, which tells you where that recognition comes from.
What we found
The two measures come apart completely. Every company below is anonymised and labelled arbitrarily, and all of them compete in the same business software categories.
Vendor A is the most quoted company in the whole study. Its website was cited 8,154 times, more than any other. And only 5.6% of its mentions survive when you remove its own website from the answer. Take their site away and the brand very nearly disappears from the conversation.
Vendor C is the mirror image. Its website was cited 479 times, roughly one seventeenth as often. Yet it was named 1,226 times, and 74.4% of those mentions happened without its own site being quoted at all. The engine names it because it knows the category, not because it is reading their marketing.
The reason shows up in one more number. 141 pages across 49 different websites mention Vendor C. Vendor B, at the bottom of the chart, has none at all. Not one other website in the entire dataset names them.
Why this happens
An AI engine builds two different kinds of knowledge. There is what it retrieves at the moment of answering, by reading pages. And there is what it already understands about a category, assembled from how the whole web discusses it.
Publishing well gets you into the first kind. It puts strong pages in front of the engine and earns citations when your page is the best available source. That is real and worth having.
Only other people talking about you gets you into the second. When review sites, industry publications, comparison articles and community discussions name you as one of the options, the engine learns that you belong in the category. After that it can name you without reading anything of yours.
The first is rented. It depends on being the best retrievable page each time. The second is closer to owned, because it persists whether or not your page wins that particular answer.
What this means for your business
The immediate implication is that a high citation count can conceal a weak position. If nearly all of your mentions arrive attached to your own website, your visibility is entirely dependent on your pages winning retrieval every single time. The day a competitor publishes a better page on a question, you do not slip down a ranking. You vanish from that answer.
The diagnostic is simple enough to run on your own reporting. Take your mentions and ask how many happened without your own domain being cited. If that share is small, you have publishing strength and no category standing.
Fixing it is not a content problem, which is why content teams tend to make little progress on it. The lever is being present in the places that discuss your category: review and comparison sites, industry publications, analyst commentary, community discussion. Those are the pages that teach the engine you exist. Our data shows a brand with 141 such pages being named constantly on a fraction of the citations of a brand with almost none.
It is also the more durable investment. Pages you publish can be outcompeted next quarter. A category position built from dozens of independent sources is far harder for a competitor to displace, because they cannot rewrite somebody else's page.
For most businesses the honest conclusion is that both matter and only one is being worked on. Publishing wins you answers now. Being discussed wins you the ones where your page is not the best source, which over a year is most of them.
Separating owned visibility from earned authority is the first cut we take at Novastacks.
Methodology
89 questions agreed with chief executives and business owners as ones their companies need to monitor, put to ChatGPT, Google AI Overview and Perplexity every day through August 2026, producing 6,601 answers and 61,536 citations.
For each answer we recorded both the companies named in the text and the websites cited as sources, which allows every mention to be classified as arriving with or without a citation to that brand's own domain. The analysis covers the seventeen companies tracked by name in this study, so it describes that named set rather than an entire market. Brand recognition is also shaped by company age, category history and offline presence, which this measure does not separate out. All brands are anonymised.
Frequently Asked Questions
What is the difference between being cited and being known by AI?
Being cited means AI quoted a page on your website. Being known means AI names you from what it understands about your category, without needing to read your site. The second survives when your page is not the best source.
How do you measure earned brand recognition in AI answers?
Count the mentions where the brand's own website was not among the cited sources, then express that as a share of all its mentions. In our study that ranged from 4.0% to 95.0% across companies.
Can a brand have high AI citations and weak authority?
Yes, and it is common. The most cited company in our study earned only 5.6% of its mentions without its own website being quoted, so its visibility depends entirely on its pages winning retrieval each time.
How do you build brand authority with AI engines?
Through third party coverage. The company with the strongest earned recognition in our data was named on 141 pages across 49 other websites. The weakest had none at all.
How was this measured?
For each of 6,601 answers we recorded both the companies named in the text and the websites cited as sources, which lets every mention be classified as arriving with or without a citation to that brand's own domain.
Do AI Engines Know Your Brand, or Just Your Website?
We separate the mentions you earn from your own pages from the ones the industry earns for you, so you can see which kind of visibility you actually have.