ChatGPT Brand Mentions
It is a fair question, and one that comes up in almost every marketing conversation right now. For twenty years, the link graph was the backbone of how Google ranked the web. So when ChatGPT, Perplexity, Gemini, Copilot and other AI answer engines start naming brands in their answers, the natural assumption is that they must be doing the same thing under the hood: counting backlinks, weighing authority, ranking pages. The honest answer is more interesting, and it changes how you should spend your effort.
Short version: AI answer engines do not use backlinks the way Google's classic algorithm does. They learn from text. And in that text, what they pay attention to most is how often, how consistently, and in what company your brand is mentioned, whether or not those mentions carry a link.
How a search engine and a language model actually differ
Google's original PageRank model treats a link as a vote. A page that earns many links from trusted pages is judged more authoritative, and that authority flows through the link graph. The hyperlink is the unit of currency. This is why an entire industry grew up around acquiring links.
A large language model works on a completely different principle. It is trained on an enormous body of text, much of it drawn from public web crawls like Common Crawl. During training, the model is not following links and tallying votes. It is learning statistical patterns in language: which words, names, and concepts tend to appear near each other. When the phrase "project management tool" shows up next to a particular brand name thousands of times across many different sources, the model learns to associate the two. That association is what surfaces later when someone asks the model for a recommendation.
The key consequence is that a brand name written in plain text, with no link attached, is fully visible to the model. The link was never the thing being counted. The text was. This is the heart of why mentions and links behave so differently across the two systems, and it is a point worth sitting with before you plan any strategy.
What the research shows about mentions versus links
The clearest public data on this comes from BrightEdge, which analyzed how AI engines actually name brands in their answers. Their finding: ChatGPT mentions brands about 3.2 times more often than it cites them, averaging roughly 2.4 brand mentions per prompt against only about 0.74 citations per prompt (BrightEdge). In other words, the model is far more willing to say a brand's name than to attach a clickable source to it. The mention is the dominant unit of visibility, not the link.
BrightEdge also found that commercial intent matters a great deal. Queries with transactional language, the kind people use when they are close to buying, drove several times more brand mentions than purely informational queries. That tells you the moments where being named pays off most are exactly the moments that matter commercially.
This shift is now widely recognized across the industry. Search Engine Land has described it as a move "from links to brand signals," arguing that the authority model itself is changing as AI answers become a primary surface (Search Engine Land). Agencies have reached similar conclusions independently; iProspect framed it bluntly as moving beyond backlinks because brand mentions matter more in the age of LLMs (iProspect), and GoDataFeed put it even more directly: your backlinks do not matter to LLMs, your brand mentions do (GoDataFeed).
So do backlinks matter at all?
It would be dishonest to say links are worthless. They retain some indirect value. Pages with strong link profiles tend to rank well in classic search, and several AI engines retrieve live search results at query time to ground their answers. A page that ranks well is more likely to be pulled into that retrieval step. A page that gets linked is also more likely to be crawled, copied, and republished, which means it appears more often in the text the model learns from.
But notice that in every one of those cases, the value of the link is upstream of the thing that actually moves the needle, which is your brand name appearing in trusted text. The link is a means, sometimes, to that end. It is not the end itself. If you can get your brand named in trusted content directly, you have skipped to the result.
What is genuinely not known
It is important to be candid here. The internal ranking and selection logic of ChatGPT, Gemini, Perplexity and the rest is not public. No one outside those companies can tell you the exact weighting of any single signal, and that weighting almost certainly differs from engine to engine and changes over time. Anyone who claims a precise formula is guessing.
What we can observe is the output, and the patterns in that output are consistent: brands that are mentioned often, across many independent and credible sources, in clear association with their category, show up in answers far more than brands that are not. That observation is durable even though the internals are opaque. Build your strategy on the observable pattern, not on a claimed formula.
What this means for your strategy
The practical takeaway is a reordering of priorities, not a complete reinvention. A few things follow directly from how these systems work:
- Earn unlinked mentions across trusted sources. The model reads the text. A clear, natural mention of your brand next to your category, on a site the model already trusts, is exactly the signal it learns from. The link is optional.
- Consistency beats any single placement. A model builds confidence from repetition. The same association, described the same way, appearing across many independent sources, is what hardens into a recommendation.
- Anchor your brand to its category in plain language. "Acme is a payroll tool for small agencies" teaches the model what you are. Vague, link-only references teach it much less.
- Do not abandon classic SEO. Strong content still feeds retrieval and still gets you crawled. It is one input now, not the whole game.
The mental model worth carrying away is simple. Google asked who links to you. AI answer engines, in effect, ask who talks about you, and in what terms. Those are different questions, and the second one is answered in plain text. That is precisely why placing natural, unlinked brand mentions across a network of trusted sources is a sound way to build the kind of presence these engines reward. If you want to see how a structured mention program is priced and scoped, take a look at our packages, or read how ChatGPT picks brands. For the wider links-versus-mentions argument, our mentions vs SEO breakdown goes deeper.