For a decade, digital PR had one currency: the followed backlink from a high-authority domain. Coverage without a link was a consolation prize. Great placement, shame about the no-follow. Agencies reported DR scores, link counts, and not much else.
That scoreboard is now out of date. It isn't wrong, because links still matter, but it is incomplete in a way that misallocates budget. AI assistants answering "who should I use for X" don't count your backlinks. They repeat what the corpus says about you. An unlinked mention in a well-crawled article, invisible on last decade's scoreboard, is exactly the material a language model learns from and a retrieval system quotes. The currency is splitting in two, and digital PR has to price both.
What each asset actually does
Backlinks: the ranking asset
A followed link from a real publication remains the strongest single input to classic search rankings, and classic search is far from dead. It still drives the bulk of most businesses' organic traffic, and it feeds the retrieval layer AI answers are built on. Links also compound: authority earned once keeps paying across every page.
But links were always a proxy. Search engines counted them because a link was a vote of confidence that was hard to fake. The vote is what mattered, and the anchor tag was just how machines could read it.
Brand mentions: the entity asset
A mention (your name in third-party text, linked or not) does different work:
- It teaches models who you are. Language models build their picture of your business from co-occurrence: your name appearing near your category, your city, your customer type, across many independent documents. Every mention is a training example. Links are irrelevant to this, because a model reads the words either way.
- It feeds AI citations. When an assistant composes a shortlist, it draws on what retrieved documents say. "Acme is a popular payroll option for restaurants" in a roundup can surface you in a thousand AI answers without passing a drop of PageRank.
- It corroborates. Google has said for years that it uses unlinked mentions as an entity signal, and AI systems visibly weight independent corroboration over self-description. Ten sites mentioning the same facts about you beats one site linking to you, for this purpose.
The honest uncertainty, as always: nobody outside the labs knows the exact weightings, and they shift. But the direction is well-supported. Being in the record beats being linked from it when the reader is a machine composing an answer.
Where links still win
Don't let the pendulum swing into "links are dead" content, because that overshoots. Pay for links with effort, not cash (buying links remains a bad idea), when:
- You're fighting for classic rankings on commercial pages. Category pages, local results, comparison keywords: link authority still decides these daily.
- You need referral traffic that converts. A linked mention in a buyer's-guide article sends actual customers. A mention alone makes them search for you and hope.
- You're building the retrieval floor. Pages that rank get retrieved, and pages that get retrieved get cited. Link authority indirectly buys your own pages into AI answers.
Where mentions win
- Category shortlist queries. "Best X for Y" asked to an assistant is answered from roundups, forums, and coverage. Inclusion in those documents, linked or not, is the whole game.
- Entity credibility. For "is [company] legit"-type questions, volume and consistency of independent mentions is what the machine has to go on.
- Speed and scale. Mentions are simply easier to earn than followed links. Journalists who'd never link will still name you, and a distributed press release puts your name and facts on hundreds of crawled news pages in a day. The links from syndication are mostly no-follow and always were. But as mentions, dated and duplicated across news domains, they're doing exactly the work this section describes. The wire was an entity channel all along. We just graded it as a link channel and called it disappointing.
The rebalanced playbook
If you run digital PR, in-house or through an agency, the shift is in emphasis, not tooling:
- Report three numbers, not one. Linked placements, unlinked mentions, and AI-answer appearances (track them with a monthly prompt log). If your agency only reports links, they're optimizing for one of your three channels.
- Stop paying premiums to convert mentions into links. Chasing a webmaster for a link edit was worth it when the mention counted for nothing. Now the mention is banking value already, so spend the follow-up effort on the next placement.
- Standardize what gets said about you. Mentions build your entity, so the words matter. Use one consistent descriptor, "[Company], the [category] for [customer]", in every release boilerplate, bio, and pitch. Repetition across sources is precisely what machines index.
- Aim campaigns at citable formats. Data studies, surveys, and concrete claims get quoted (mentioned) even when they don't get linked. The classic digital-PR data campaign is more valuable now, not less. It just has to be measured differently.
- Keep a floor of on-the-record announcements. Earned coverage is spiky. The record shouldn't be. A steady cadence of distributed releases keeps dated facts flowing into the corpus between hits, at a cost that's a rounding error next to agency retainers.
Bottom line
Backlinks are the ranking asset, and mentions are the entity asset. You need both, but the default digital-PR scoreboard only counts the first, which means most programs are underinvesting in the input AI answers actually consume. Rebalance toward being named, described consistently, and corroborated widely, keep earning real links where rankings pay, and grade your PR by what both kinds of machines can see.
The cheapest scalable source of consistent brand mentions is the wire. [See what it costs](/pricing).