Digital PR earned its budget on a simple chain: a story worth covering earns coverage, coverage carries links, links move rankings. The chain still works — nothing in this article retires it. What has changed is who reads the coverage. For twenty years the only machine audience that mattered was a crawler counting links and weighing the domains they came from. Now the same coverage is read by models — systems that assemble answers, cross-check claims and decide which brands are safe to name. That second reader values different things in the same placements, and a PR program designed only for the first reader leaves most of the new value on the table.
What digital PR was built to earn
Strip the campaign creative away and digital PR was always an authority acquisition discipline. Search engines treated an editorial link as a vote — an independent publication choosing to reference a site was evidence the site deserved to rank — and the votes were weighted by the credibility of the voter. Earned media happened to be the cleanest legitimate way to collect those votes, so an industry formed around engineering it: campaigns designed to be coverable, data angles built for headlines, outreach lists of journalists sorted by the domain authority of their mastheads.
The economics followed the metric. Success was reported in linking root domains; a placement that carried no link was a consolation prize; brand awareness was described, honestly enough, as a by-product. The coverage was never really the product. The link inside it was.
The coverage now has a second reader
AI engines change the audience for that coverage without changing the coverage itself. When a model grounds an answer about a category — who the credible providers are, what a brand does, whether a claim holds — it is not counting links. It is retrieving text from sources it trusts and cross-referencing what those sources say. Independent coverage functions as corroboration: third-party evidence that the brand exists as described, does what it says, and is spoken about by people who do not work for it.
This is the corroboration leg of entity building. Engines assemble an identity for a brand from two layers — what the brand declares about itself and what everyone else says about it — and they believe the first layer only to the extent the second confirms it, a mechanism covered properly in the entity SEO article. It is also the substance of the authority factor in the citation framework: a fact that appears only on a brand’s own site is a fact a model has no way to check, and grounded systems are built to avoid repeating unverifiable claims. Earned coverage is where a brand’s claims stop being self-asserted.
Which quietly rehabilitates the unlinked mention. Under link logic, a journalist naming a brand without linking to it delivered close to nothing. Under citation logic, an accurate description of the brand in a publication the model trusts is exactly the corroborating evidence the grounding process looks for — link or no link.
Links are ranking currency; mentions are citation currency
It helps to think of two economies running in parallel over the same activity.
Links remain the currency of rankings. Crawlers still follow them, algorithms still weigh them, and organic positions still respond to them. Any PR program that abandons links entirely is giving up a compounding asset that continues to pay.
Mentions — accurate, independent, attributable descriptions of the brand — are the currency of citations. When an engine decides which businesses to name in a generated answer, the deciding evidence is not the anchor text pointing at the brand’s site; it is whether credible third parties describe the brand consistently, attach real people to its expertise, and repeat its key facts without contradiction.
The practical consequence lands in the negotiation every PR practitioner knows: the follow-up email asking the journalist to add a link. That email made sense when the link was the entire value of the placement. It makes less sense now. An accurate, well-attributed description in a masthead the engines trust can be worth more to a brand’s citation position than a followed link from a site they barely weigh — and a placement should be valued on both currencies before anyone decides it failed.
Quality beats volume, harder than ever
Link economics always favoured quality, but they tolerated volume, because links summed. A hundred modest links still added to something, and an industry of volume tactics — mass outreach, reactive commentary at scale, syndication networks — existed because the sum was real.
Corroboration does not sum the same way. A named journalist at a credible publication, independently choosing to describe a brand and its work, is a genuine second voice — provenance a model can weigh. Fifty syndicated reprints of the same wire release are one voice, echoed. The text is duplicated, the byline is nobody’s, and retrieval systems are engineered to deduplicate near-identical content, and tend to discount sources with no editorial identity. Syndication multiplies links; it does not multiply corroboration.
The same logic downgrades the churn end of digital PR — the listicle placements, the paid-inclusion roundups, the guest posts on sites that publish anyone. These carried marginal link value in the old economy. In the new one they carry approximately nothing, because the reader now evaluating the coverage is precisely a system built to ask: who wrote this, for whom, and would they say it if the brand had not asked? Five placements a year that survive that question outperform fifty that do not — on citations immediately, and increasingly on rankings too, as the engines’ quality systems converge on the same judgement.
Original data is the strongest PR asset
If corroboration is the prize, the strongest move available is to become the thing others corroborate — and the reliable way to do that is original data.
A brand that publishes a genuine finding — a measured result, a category benchmark, an index built from its own records — changes its role in the coverage economy. It stops asking journalists for attention and starts supplying them with something scarce: a fact that did not exist before. Coverage of a data story cites the source by necessity, describes the brand accurately by necessity, and attaches the finding to the brand’s name in exactly the durable, cross-referenced form that answer engines inherit. This is the logic visible across the industry’s own research into AI citation behaviour: the firms that measured how engines choose sources became, by publishing the measurement, sources the engines choose. It is also the logic behind this site’s own research program — publish the study you want to be quoted from.
The data does not need to be large to be original. An SME’s own operational records — anonymised, honestly framed, with the method stated — can yield a finding no publisher can get elsewhere. What it does need is integrity: a stated sample, a stated period, and no rounding-up for the headline. A data story that collapses under one journalist’s scrutiny costs more than it earned, because the correction enters the corpus alongside the claim.
Anchor text fades; factual consistency rises
The old technical obsession of digital PR was anchor text — the clickable words in the link, negotiated and optimised because exact-match anchors once moved rankings directly. That lever has been depreciating for years, and in the citation economy it barely registers. Models do not care what words carried the link. They care whether the facts in the coverage match the facts everywhere else.
This replaces one obsession with a more demanding one. Engines cross-check a brand’s story across its site, its profiles and its press coverage, and treat agreement as settled fact and conflict as a reason to hedge or go silent. Coverage is part of that record — permanently. A spokesperson who tells a journalist the firm operates in eight cities while the website says six has not rounded up; they have planted a contradiction in the corpus the brand will be checked against for years. Every piece of coverage is now a permanent entry in that corpus, which means the quote in the article has to match the facts on the site — same company descriptor, same service language, same numbers, same claims.
Handled deliberately, this cuts the other way and becomes an asset: every placement is an opportunity to lay down one more independent copy of the brand’s canonical facts, in a source the engines trust more than the brand itself.
A program shape for an SME
None of this requires an enterprise budget. It requires different selection criteria and one internal document most businesses have never written.
Fewer, better placements. Set targets in credible-outlet placements with named journalists, not in linking domains. A realistic quarter for an SME is one to three placements that a model would treat as independent corroboration. Decline the churn even when it is free — it dilutes nothing except your own signal, but it spends the team’s capacity on coverage the second reader ignores.
One real data story per year, minimum. Build it from data the business actually owns, state the method, and let it anchor the year’s outreach. A single finding that trade press cites outworks a year of commentary requests, and it keeps paying as the coverage enters the retrieval corpus.
A consistency brief for spokespeople. One page: canonical company name, the category noun the business commits to, locations, the approved numbers, and the exact phrasing of the three or four claims the brand wants repeated. Every interview, byline and podcast appearance follows it. This is the cheapest item in the program and the one that most directly converts coverage into citation-grade corroboration.
Measure both currencies. Track links as before, and alongside them track whether the engines now cite and describe the brand when asked the buying questions — the method in the self-audit guide covers it in an afternoon a quarter.
The discipline, restated
Digital PR is not being replaced; it is being re-priced. The placements that were always genuinely earned — real journalists, real publications, real stories — are worth more than they were, because they now pay in two currencies at once. The placements that were only ever link-shaped are worth less, and the gap will keep widening as answers displace lists. Reading a brand’s existing coverage the way a model reads it — what corroborates, what contradicts, what the engines actually retrieve — and shaping the PR brief that follows is part of the work of AI search optimisation. The brands that adjust the brief now will spend the next few years being quoted on facts they placed deliberately, in words they chose.