For twenty years, link building rested on a simple equation: the more quality links a site earned, the higher it climbed in Google. A link was a vote, and we counted the votes. In 2026, a second layer has been laid over the first. When someone asks ChatGPT, Perplexity or Google's generative answers a question, they no longer see ten blue links: they see an answer, with a handful of cited sources. The real question is no longer just "how do I earn links", but "how do I become one of the sources the AI chooses to cite".
Good news: link building is not dying. It is changing nature. The link still passes authority for classic organic ranking, which remains the raw material generative engines draw from. But a second signal, long underrated, is taking the lead for AI: the brand mention, with or without a hyperlink. This article bridges classic link building and AI visibility: what I call the link building × GEO bridge.
A link lifts a page. A mention makes a brand exist in the model's memory. In 2026, both count, but it is the second that decides whether the AI cites you or your competitor.
From counting links to entity signals
A language model does not rank pages: it builds a representation of your entity (your brand, your product, your expert) from everything said about it on the web. That representation feeds not only on clickable links, but on every place your name appears in a relevant context. It is a paradigm shift: during the PageRank era, an unlinked mention was a weak, almost lost signal. In the age of AI, it is a front-rank signal.
The figures circulating since early 2026 are telling. Several analyses converge on one point: the frequency of brand mentions correlates roughly three times more strongly with AI citation rates than the number of backlinks (on the order of 0.66 versus 0.22 in correlation coefficient). In other words, a brand named often across recognized sources, even without a link, is picked up far more often than a brand that has accumulated links but that no one talks about. The vote has changed shape: we no longer just count links, we weigh presence.
How ChatGPT and Perplexity choose their sources
The two tools do not work the same way, and you need to understand that to act. Perplexity queries the web in real time: for a query it retrieves a handful of pages (often five to ten) and cites only three or four. Its selection combines several criteria: relevance, recency, entity clarity, content extractability, source authority, attribution quality. A poorly structured page, even one that ranks well, can be dropped in favor of a source the machine finds more "readable".
ChatGPT relies first on its training corpus, frozen at a cutoff date, and only activates web search when needed. When it cites, it favors high-authority sources, Wikipedia foremost. This has a direct strategic consequence: being present in the major reference sources (encyclopedias, established media, the platforms the AI consults such as Reddit or LinkedIn depending on the topic) weighs as much as, if not more than, accumulating links on second-tier sites. I covered the community-platform case in my article on Reddit, the source AI cites.
The link does not disappear: it becomes a foundation
It would be wrong to conclude that link building should be abandoned. Generative engines, including Google's answers, overwhelmingly start from pages that already rank well in the organic results. And that ranking still depends on links. The quality backlink therefore remains the foundation: it gets your pages into the pool the AI draws from. What I evolve for my clients is not the end of links, but their purpose.
Concretely, I still target contextual editorial links (the ones that live inside real content, on a relevant site) as I explain in my article on small editorial links. But I now assess them against a second criterion: does this publication, on top of placing a link, name the brand in a clear context? A bare link at the bottom of a directory brings almost nothing to the AI. An in-depth article that cites your expert, describes your method and places a link feeds both ranking and entity representation. Same effort, two benefits.
The real 2026 lever: digital PR
If one channel dominates for AI visibility, it is digital PR, earned media. 2026 studies estimate that the vast majority of AI-produced citations point to third-party editorial content rather than to pages owned by the brand itself. The logic is mechanical: a brand cited across several recognized publications sends a legitimacy signal that no "owned" page can send. And source diversity counts as much as source count: being picked up by ten different outlets weighs more than ten mentions on the same one.
That is why a modern link building strategy increasingly looks like a public relations strategy: identify the topics where your expertise has something unique to say, produce genuinely citable data or viewpoints, and get them picked up by the publications your audience, and the AI, consult. The goal is no longer just the link: it is the repeated, diverse and contextual mention of your brand as a reference on a subject.
| Signal | Classic SEO effect | AI citation effect |
|---|---|---|
| Contextual editorial backlink | Strong (authority, ranking) | Indirect (gets the page into the pool) |
| Unlinked brand mention | Weak | Strong (entity representation) |
| Pickup by several distinct outlets | Medium | Very strong (source diversity) |
| Wikipedia / major-reference presence | Medium | Very strong (source favored by ChatGPT) |
| Bare directory or footer link | Weak to none | None |
| Structured, extractable content | Medium | Strong (makes citation easier) |
The costliest mistake
Keep buying links by the volume, on sites with no audience and no editorial line, hoping to "boost the AI". Those links already bring little to ranking, and strictly nothing to your representation in the models: no one is talking about you, they merely link to you. Conversely, a single pickup in a recognized outlet that names your brand and describes what you do is worth more, in an AI's eyes, than a hundred anonymous links. The right reflex: judge every link building action by the question "does this make people talk about the brand, or merely point at it?".
Building a cited presence: the method
The approach I apply comes in four steps. First, consolidate the entity: a clean "About" page, structured data, a coherent record (name, expertise, contact) everywhere the brand appears, ideally a presence on the major references. This is the prerequisite for the AI to know who you are before deciding to cite you: work I tie to my GEO expertise and, where admissible, to Wikipedia presence.
Next, produce citable material: clean data, quantified field feedback, sharp viewpoints. The kind of content a journalist picks up and a model extracts readily. Then distribute: digital PR, op-eds, appearances, to multiply mentions across diverse, trustworthy sources. Finally, measure differently: track not only rankings and links, but share of voice in AI answers, citation frequency and associated sentiment. This logic of structure, authority and internal linking ties directly to my SEO expertise: link building is no longer an end in itself, it is a lever serving a presence that engines, classic and generative alike, judge worth citing.
In the crucible: turning the link into a citation
Link building has not been made obsolete by AI; it has been put back in its place. The link remains the foundation that makes your pages exist in the organic results, the raw material of generative engines. But the noble metal, now, is the mention: the brand named, recognized, picked up, until it becomes an obvious answer for the model to cite. The lead, here, is the anonymous link bought by the volume; the gold is the contextual citation that carries authority.
If you want ChatGPT and Perplexity to cite you rather than your competitors, the challenge is to stop counting links and start building a presence people talk about. I tie together entity consolidation, citable content, digital PR and editorial link building in a single profitability logic on my SEO expertise page, so your brand moves from being a link to being a cited reference.
Become a source the AI cites
I audit your presence (link authority, mention frequency, entity representation, citability) and hand you a plan: which links to target, which mentions to trigger, and how to measure your share of voice in AI answers.