Press release distribution is an old service, and most of what it once promised no longer happens: the journalist rarely calls, and the reader rarely scrolls. What a distributed release still does — put a dated, third-party record of you on domains you do not control — has become more valuable, not less, because AI answers are assembled from exactly that kind of record.
eigenKOR writes releases to be read by people and by AI crawlers alike, distributes them into the news networks the engines index, and measures whether the answers changed. This page explains what that means, what we do, and what we will not do.
A conventional press release was written for one reader: a journalist deciding whether to cover a story. That reader still exists, but is no longer the one who decides whether a buyer hears your name. Three others do, and each wants something different from the same document.
| The reader | What it takes from a release |
|---|---|
| A journalist | A story worth a reporter's time. Most releases are not one, and the ones that are earn coverage slowly and unpredictably. We write for this reader honestly, and we do not promise them. |
| A buyer searching your name | A third-party record. Something on a domain that is not yours, dated, saying who you are and what you did — the difference between a company that says so and a company somebody else wrote about. |
| A search engine | A real page at a real address, indexed, on a domain with standing, that agrees with the other pages about you. Links and mentions on other people's domains are still where search looks for corroboration. |
| An AI answering a question | A passage it can lift whole: one that names you inside it, states one specific fact, makes sense with nothing around it, and sits on a domain the system already trusts. Then a second domain saying the same thing. |
A release written for the first reader alone is, today, a relic. Written for all four, it is source material. The craft of writing for the fourth reader is what this page is about.
When a buyer asks ChatGPT, Gemini, Claude, Copilot, Perplexity or Google's own AI who to hire, the answer is not built from your website. It is assembled from passages retrieved from sources the system trusts, about entities it can identify with confidence. Your own page is treated for what it is: a claim you make about yourself. The same fact on a news domain you do not own is a third party repeating it, and agreement between independent sources is what a retrieval system weighs when it decides which names to use.
The industry has its own word for what a distributed release produces: a pickup — your release republished, wording unchanged, on a news site you do not control. For a human reader that ranks below a written article, because no reporter interpreted it. For a machine the order inverts. A rewrite loses your wording; a pickup keeps it, which means one identical, controlled description of you now sits on many domains at once, in a format that chunks and attributes cleanly.
Press release distribution is the established category, and we do that work. A release is written, published on your own newsroom, and syndicated into news networks, where it produces indexed pages on domains you do not own. That is what a buyer types into a search box, and it is where the work on this page begins.
What we add is the engineering and the measurement. The release is written for retrieval: one claim per passage, the entity named in every unit, specifics rather than adjectives, structured data on the version we control, and a boilerplate that is identical everywhere it appears. It is distributed in an order and to surfaces chosen for how AI systems index them, not for how many outlets a report can list. And what the AI says when a buyer asks is measured before the first release and again afterwards, on the same questions, against control questions we deliberately never target. We call the document an Answer-Engineered Release, and the loop around it the Citation Engine.
Distribution firms count outlets. PR firms pitch journalists. SEO firms count links. The evidence an AI reads is fed by all three, and it is rarely anybody's job to make the record agree with itself and then check what the machine concluded. That is the job this page describes.
See what the machine says today. The questions your buyers ask AI about your category, run across every engine we can reach, several times each, with the sources each answer cited. The same read on your closest competitors. We baseline before anything is written, and you keep the baseline whatever you decide next.
Build the question set. AI prompts are not keywords; they are long, comparative and conversational. We source them from your sales conversations, your support record and your category's own discussions, and hold the set fixed so that a before and an after mean something.
Decide what is worth being cited for. True, specific, attributable claims — a number, a date, a named source, a defined mechanism — each traced to a document you can produce, each written in the shape of an answer to a question a buyer actually asks.
Write the release for retrieval. Headline as a declarative answer, a lead paragraph that stands alone, claim blocks that each carry one fact and your name, quotations that carry a claim rather than a feeling, a question-and-answer block, the canonical boilerplate, and the structured data that ties it to you.
Publish in order. Your own newsroom first, so the canonical record exists; then the news networks the engines index; then the reference and directory surfaces that resolve your identity; then editorial and community work over the following weeks, where it is genuinely earned.
Ask the same questions again. Same prompts, same method, same engines, against the controls. Which answers named you, which domains they drew on, and whether the movement was ours or the category's. Every report says which engines were covered and which could not be.
The canonical version, with the full structured data, published before anything else so that every syndicated copy has a record to corroborate. A release that syndicates before its home exists produces an entity the systems cannot resolve.
Syndication into networks selected for how AI systems index and cite them, not for the length of their outlet list. The release appears at real addresses, in the news-release sections of real news domains you do not own.
Industry databases, professional directories, credential registries, and the open knowledge bases where you genuinely qualify. This is the backbone of how a system decides who you are, and the most neglected surface in the market.
Where the story earns it, outreach to trade publications and industry newsletters. Slower, never promised, and higher in trust than any syndicated placement. It is scoped by the quarter, not the month.
Where your customers discuss the category, helping your own people show up accurately and openly. We do not post as anyone we are not, and we do not manufacture a conversation.
The facts the release states are the facts your website, your business listings and your profiles state. A release that contradicts the rest of your record teaches the system to trust none of it.
The distribution trade has a thin end, and platforms have spent a decade learning to discount it. So the boundary is stated here, in full.
This is not only a preference. Fabricated testimonials and endorsements are prohibited in the United States by a federal rule that came into force in October 2024,2, and a release built to pass ranking credit through optimised links is a pattern search engines have discounted for years.1 A firm that offers either is offering you a liability with your name on it — and a record the machines have already learned to ignore.
| What happened | What the release carries |
|---|---|
| A launch or a new service | What it is, who it is for, what it replaces or changes, and the one specific a buyer would want to check — written as the answer to the question they would ask about it. |
| A hire or an appointment | The person, resolved: name, title, prior record, the profiles that identify them. A leadership release is entity work for two entities at once. |
| A milestone or a data finding | The number, the date, the method. Findings from your own operation are the most citable claims a business owns, and most businesses never publish them. |
| A partnership or an expansion | Both parties named in every passage, the geography stated plainly, and the facts agreed with the partner before publication so that two records do not disagree. |
| An award, a credential or a certification | Who granted it, when, on what criteria. Third-party recognition is only evidence if the third party can be identified. |
| Nothing this month | No release. We turn to the work that does not need news: listing consistency, structured data, identifier profiles, the boilerplate sweep. A release with nothing to say is the thin content platforms discount, and it spends your credibility to publish it. |
A single release, measured before and after, is an experiment: a controlled description of you, placed, then checked against what the machine said before it existed. It is a fair test and we run it as one. But the systems that answer buyers weigh recency, and they weigh agreement between sources, and neither is built by one document. What builds them is a record that keeps arriving — each genuine development written for retrieval, published in order, and checked against the last one so that nothing about you disagrees with anything else about you.
That is why this is a practice rather than a purchase. Releases when there is news, infrastructure work when there is not, and the same questions asked of the machine at every interval so that the record's effect is measured rather than assumed. A business with that record can absorb a stale listing, a competitor's launch or a platform change. A business with one press release from three years ago cannot, and has no way of knowing.
Release writing, syndication, structured data, business listings, entity and identifier work, editorial outreach and competitive measurement are, in most agencies, separate departments or separate vendors. The evidence an AI reads does not respect those boundaries. It reads the whole record, and a release that says one thing while your listings say another is a gap in the summary, not a placement.
Doing the work under one roof means the release, the newsroom page, the directory entries, the structured data and the measurement are built to agree with each other — which is the whole point, because agreement between sources is what the machine half of your visibility is made of.
Nobody controls the output of somebody else's system, and the same question asked twice can return two different lists. We can make the record it reads accurate, consistent and abundant, and we can measure what it says before and after. Anyone offering a guaranteed citation is selling something they do not have.
Named destinations invite a comparison of lists, imply editorial coverage where there is syndication, and go stale the day a network changes. We describe the mechanism, show you the placements that published, and report whether the answers moved.
How long this takes is set by the state of your record before we start, which nobody can know before measuring it. Some engines move sooner than others, and we will tell you the order. We will not tell you the week.
What this costs depends on how much of a record already exists, how many surfaces disagree, how much genuine news you generate, and whether your field's advertising rules require a compliance review of every claim.
Editorial coverage is commissioned effort, not purchased inventory, and a firm that guarantees it in a named publication is describing something no honest newsroom would agree to. Where a story earns coverage we pursue it, and we tell you when it does not.
If you have nothing true to announce and no data of your own to publish, a release is not your lever yet, and we will say so; the work that comes first is the record on your own site and your listings. If what you want is a count of backlinks, a wall of outlet logos or a story placed in a named publication by a given date, we are not the firm, and no honest one is.
If you practise in a regulated field — law, medicine, finance, health products, real estate — this work is available to you, and every claim in every release is approved in writing by your own compliance function before it ships. We do not substitute our judgement for theirs, and no release goes out without that approval on file.
A press release written and published for the readers that now decide whether a business is found: a buyer searching its name, a search engine indexing the page, and an AI assembling an answer. It is published first on the business's own newsroom with structured data, then syndicated into news networks, so that a dated, third-party record of the business exists on domains it does not control. Done well, it is evidence. Done badly, it is a keyword page with a dateline.
A traditional service sells distribution: the release is sent to a list of outlets and the report counts them. We write the release differently — one claim per passage, the business named in every unit, specifics that can be checked, structured data on the canonical version — distribute it in a deliberate order to surfaces chosen for how AI systems index them, and then measure what the AI says against a baseline and against control questions. The distribution is part of the work. It is not the work.
Nobody can promise that, and we do not. What we can do is make the record those systems read accurate, consistent, specific and present on the domains they draw from, and then measure whether the answers to your buyers' questions changed. We report the result including when it did not move, and we hold out questions we never target so that movement can be told from category drift.
We do not name destinations in advance, for three reasons. A list of outlets invites a comparison of lists, which is the wrong measure. A syndicated placement appears in the news-release section of an affiliated site, not as an editorial feature, and a named outlet implies that it does. And networks change, so any list we printed would be stale. After publication you see exactly where the release appeared, at which addresses, and which of those domains the AI in your category already cites.
As a by-product, yes: a release produces indexed pages on real news domains, with mentions of your business and links to your site, and search engines still look to other domains for corroboration. But a release built to pass ranking credit through keyword-rich links is what Google's spam policy calls link spam,1, and we do not write them that way. The search value follows from the record being real. It is not the reason to publish.
Whenever there is a genuine development — a launch, an appointment, a finding from your own data, a partnership, a credential — and never when there is not. Recency is weighed by the systems that answer buyers, so a record that keeps arriving does more than one that arrived once. In a month with no news, the work turns to the parts of your record that do not need any: listings, structured data, identifiers, the boilerplate everywhere it appears.
Time with whoever owns your sales conversations, because the questions buyers ask people are the questions they now ask machines. The documents behind every claim we will make, because no claim ships without a source. Your approval in writing on each release, and your compliance function's approval if your field requires one. And a place on your own website where the canonical version can live at a permanent address.