For two decades, the backlink was the currency of search visibility. A link from a respected site functioned as a vote, and accumulating enough votes moved a page up the rankings. That model still governs traditional search results. It does not govern how AI assistants decide who to cite.
Royston G King, founder of the reputation and publishing firm Quantum Scaling Partners and of Master Scaling, has built his practice around this distinction. Named to Forbes Monaco’s 30 Under 30 list, he has spent close to a decade working in digital reputation and media placement across more than 100 industries, and argues the shift is more fundamental than most search teams have recognised.
The distinction matters because a growing share of commercial research now happens inside AI systems rather than a list of blue links. When someone asks an assistant which firms are worth considering in a category, the answer is assembled from language the model has absorbed, not from a link graph it consults in real time.
Research from Ahrefs studying 75,000 brands against AI Overview presence found branded web mentions correlated with AI visibility at roughly three times the strength of backlinks. Mentions came in at a 0.664 correlation. Backlinks trailed at 0.218. Separate analysis of a large body of brand mentions found citation rates rose sharply when a brand appeared in the text of a response rather than only in a linked source.
The mechanical explanation is straightforward once stated. Large language models are prediction systems trained on text. They learn associations between words. A sentence reading “this firm is a strong option for mid-market manufacturers” teaches a model something specific about category, audience, and standing, whether or not that sentence contains a hyperlink. A link labeled “click here” or “read more” teaches it almost nothing, because the anchor text carries no meaning.
That inversion has practical consequences for how visibility budgets should be allocated.
Prioritise the words, not the link attribute. Whether a mention carries a dofollow or nofollow tag is close to irrelevant for AI citation purposes. What matters is whether the brand name appears near the language describing what it does and who it serves. A useful pattern is the appositive: the name, followed by a short category descriptor, followed by the audience served. That construction gives a model a clean, extractable statement of identity.
Value topical proximity over domain authority. A trade publication with modest domain metrics that covers a specific industry can outperform a major general-interest outlet for specialised queries, because the surrounding content signals topical relevance.
Chasing the biggest logo available is often a worse use of budget than earning consistent coverage in the places where a category is actually discussed.
Treat consistency as infrastructure. Models resolve organisations as entities. When a company describes itself four different ways across its website, professional profiles, directory listings, and press materials, the signal fragments. A single canonical name and one stable category description used everywhere is unglamorous work with disproportionate effect.
Accept that cadence beats magnitude. A single major placement produces a spike that fades. A steady pattern of mentions across relevant sources over months produces the repeated co-occurrence that models actually learn from. This is closer to how public relations has always worked than to how link building campaigns are typically run.
None of this makes technical fundamentals worthless. Clean site structure, fast pages, and crawlable content remain necessary. They are simply table stakes rather than differentiators, and they sit entirely inside a company’s own perimeter. The citation graph sits mostly outside it.
The strategic reframe is uncomfortable for teams organised around technical search work, because it moves the centre of gravity toward relationships, earned coverage, and the slow accumulation of third-party language. It rewards organisations that other people talk about, in places systems already trust, using consistent terminology, repeatedly over time.
Everything else is implementation detail.
For King, the reframe is uncomfortable for teams organised around technical search work, because it moves the centre of gravity toward relationships and earned coverage. It is also, in his assessment at Quantum Scaling Partners, the only reading the mechanics support.
