Press releases have the worst reputation of any tactic in link building, and they earned it. For about a decade they were sold as a link product: pay a wire, get several hundred syndicated copies of the same optimised text, count the referring domains, invoice the client. Search engines took that apart, correctly.
Bradley Benner is making a different argument, and it is worth separating from the old one before you dismiss it. His case for the format has nothing to do with the links.
The Old Reason Press Releases Failed
The failure mode was structural. Syndication produces near-identical copies across domains that have no editorial relationship with the subject, which means the “links” were never editorial decisions. Nobody at those properties chose to reference you. The wire did, once, on your instruction, and then a distribution network copied it.
That is a scheme by any reasonable definition, and treating it as one is not a Google quirk. So the corrective advice — do not buy press releases for links — was right, and remains right. What has changed is that links are no longer the only way a mention pays.
Benner On Freshness And AI Visibility
Benner runs Semantic Links, a white-label off-page agency, alongside Semantic Mastery. On Unscripted SEO he framed the whole off-page category around citations rather than links. In his terms a citation is a brand mention plus one corroborating data point — name and website, name and phone, name and address — and it does not have to be a directory listing to count.
His words: “press releases, branded assets, brand mentions on blog posts, things like that, those all can act as unstructured citations. And that’s what influences LLM or AI search visibility right now, more than anything.” On the format specifically:
Press releases work really, really good for helping to improve AI search visibility because you get brand references on high authority media and news style like freshness factor type properties that the LLMs like to cite.— Bradley Benner, Semantic Mastery / Semantic Links
He is explicit that this sits alongside on-page work rather than replacing it — he says you can and should still optimise on page to improve the chance of your own domain being referenced. The release is a way to get the brand attached to a fact somewhere that is read and re-quoted, faster than you can earn the same placement editorially.
One more thing from him worth carrying into this: he treats relevance as the governing filter across all off-page work, and describes the job as forcing the models to recognise the associations between a brand, its services, and the places it serves. A release is one more chance to state those three things in a place a machine will read.
What A Citable Release Contains
This section is our own guidance rather than Benner’s prescription. If the value is in being quotable, then the release has to be written to be quoted, which is almost the opposite of how release copy is usually produced.
- One fact in the first two sentences. Not a summary of the company. A dated, specific, checkable statement. Machines and journalists both stop reading at the same place.
- A named person with a title. An unattributed sentence cannot be safely quoted by anything. A named operations director can.
- An original number. If you have data nobody else has — response times, volumes, survey results — that is the sentence that gets lifted.
- A source for anything you did not generate. Cite the study you are referencing rather than asserting it. A claim that can be verified is a claim that can be repeated.
- Entity clarity. Brand, what it does, where it does it, spelled out plainly at least once in a full sentence.
- Boilerplate that is not stuffed. No optimised anchors, no keyword lists. The moment the release is written for anchor text you have rebuilt the tactic that failed.
Wire Tiers Worth The Money
Honest caveat first: Benner did not name specific wires, tiers or prices, and neither will we invent them. What the mechanism implies is a way to evaluate any distributor rather than a shopping list.
If the value is a brand reference on a property that gets read and re-quoted, then the only question that matters about a distributor is which properties actually publish the release as a page with your named quote intact. Not how many outlets are claimed. Not the total “potential audience” figure. Ask for example placements from the last month in your sector, open them, and check whether the quote and the attribution survived. Most of the price difference between tiers evaporates under that single check.
How To Measure Something That Is Not A Link
This is where most programmes will fail, because the reporting habit is older than the tactic. If you commission releases on this rationale and then report referring domains, you will conclude it did not work — and your report will be measuring the mechanism you were explicitly told not to buy.
- Log placements, not domains. A list of URLs where the release actually appeared as a page, with a note on whether the named quote survived.
- Track unlinked mentions of the brand over time. The release should raise the rate at which the brand name appears in text you did not publish.
- Ask the answer engines the question your customer would ask. Record what gets named, monthly, in the same wording, so the comparison means something.
- Watch branded search. Blunt, laggy, and the hardest to argue with.
None of this is a reason to move budget out of earned coverage. It is a reason to stop pretending the format has only one use. If you want the version of this argument built around earning the citation rather than buying the placement, that is digital PR for the AI consensus engine, and the local variant is earning editorial links with owner quotes.