Industries
Press releases stopped earning links and started earning citations. We build the entity, schema, and measurement layer that decides whether an answer engine attributes the story to your client.
Cited
Citation share
Earned
Earned vs syndicated
Trial
Funnel conversion
Before teams bring us in, these are the constraints holding growth back.
Wire distribution reports pickups, not outcomes. Several hundred nofollow copies sitting on syndication mirrors reads as reach in a coverage report and moves nothing the client can bank.
Releases are written for an editor, not for extraction. With no question-and-answer block, no named expert carrying a verifiable bio, and no schema on the newsroom, an assistant summarising the topic has nothing clean to attribute and cites someone else instead.
AI referral traffic arrives unattributed. GA4 reports assistant referrals as a rounding error, so the surface now shaping buyer perception is invisible in every report the agency sends its client.
For distribution platforms the funnel is the product. Signup, first release published, and trial-to-paid are three separate drop-off points, and they are almost always reported as one number.
Growth and conversion programs for public relations agencies, in-house communications teams, and the platforms that distribute their work. GrowNexus has run this scope for a US press-release distribution platform, where the brief was self-serve funnel conversion rather than coverage volume. The client is not named here under NDA.
The measurement problem in this sector changed direction in 2025. Press releases lost their link value years ago, but industry reporting through 2026 puts PR content — releases, bylines, case studies, and earned editorial — at the large majority of what AI assistants cite, with structured releases rising roughly fivefold as a citation source between July and December 2025. PR Newswire and Businesswire both now sell AI-optimised release formats and citation tracking as paid products, which is the clearest available signal that the category has already moved.
So the work is no longer pickup counts. It is entity clarity, extractable question-and-answer structure, schema on both the release and the newsroom, and a reporting line that shows whether an answer engine attributed the story to your client or to a competitor summarising the same news. We record the current citation and referral picture before changing anything, then report movement against that baseline.
The 360° stack, scoped for pr & communications operators.
Rank in Google. Get cited by ChatGPT, Gemini, and Perplexity. Own both search landscapes.
Senior growth strategy without the full-time executive price tag.
Websites and apps that load fast, rank in search, and turn visitors into paying customers.
Earn media coverage, build executive authority, and control the narrative before someone else does.
We agree the definitions and data source before work starts, then report movement without inventing a universal benchmark.
What we have measured
Published in full, with the method and the dates. Read it before you decide whether we know your market.
10 min read
Assistant referrals look like a rounding error while page-one buying questions earn no clicks. Four measurement methods, what each one misses, and the three report splits that change what you work on next.
8 min read
One checkable fact, a named human, real questions, and the right schema. What actually makes a release extractable in 2026 — including the FAQ markup change most advice has not caught up with.
8 min read
Yes, but not through links — Google's spam policies name them directly. What a release still does, what it stopped doing, and the one test worth applying before you pay to distribute anything.
8 min read
A pickup number counts sites that auto-republished your feed, not people who chose to cover you. The three tiers hidden in one figure, how to separate them in five minutes, and how to report it honestly.
10 min read
A practical 0–2 benchmark for news sites and digital publishers: check Discover eligibility, images, headlines, authorship, measurement, and revenue context without promising traffic.
10 min read
We measured thirteen competing websites for AI search readiness on 1 August 2026 and re-checked them on 12 August. One published an llms.txt, eight had no structured data, and the only site restricting an AI crawler was one we built.
12 min read
We re-analysed 18,151 cited pages across ChatGPT, Google and Perplexity. Relevance, structure and evidence were associated with deeper use, but they are not magic ranking factors.
8 min read
The llms.txt spec was revised in August 2026. We measured 25 major sites on 13 August: 18 publish a file, but only 2 have implemented the discoverability relations v2 added, and none ship them in HTML. Here is how to write one properly.
FAQ
Not the way they did. Google discounted links inside syndicated releases years ago, so distribution volume no longer builds authority. What remains is indexation, entity reinforcement, and inclusion in the corpus AI assistants draw on — which is why the wire services themselves now sell AI-optimised formats. Treat a release as a citation asset and an entity signal, not as a link buy.
Four things carry most of the weight: a factual claim stated once, plainly, near the top; an explicit question-and-answer block using the phrasing people actually ask; a named expert with a verifiable bio and a linked profile the model can resolve to a real entity; and schema on both the release and the newsroom that marks up the organisation, the person, and the questions. We then test the release against live assistants rather than assuming it worked.
Partially, and we are explicit about the limits. Referral traffic from assistants is capturable in analytics but understates reality, because most citations produce a read and no click. So the measurable layer is a tracked prompt set run on a fixed schedule across the major assistants, recording whether the client is named, how they are described, and which source the answer points to. That gives a trend line. It is not the same as a rank report and we do not present it as one.
Yes. Search, schema, and AI-citation work can be delivered under your brand, with reporting formatted for you to pass to your client directly. This is the most common shape of engagement for agencies who own the media relationships and do not want to build a technical team to sit behind them.
Both. Platform work is a different scope: it is self-serve conversion, activation, and retention rather than campaign delivery. We have run that scope for a US distribution platform, covering the path from signup to first published release to paid conversion as three separately measured steps.
Both, and they are different engagements. Client-side work is the release layer: schema, entity clarity, and citation measurement on the coverage you place. Agency-side work targets your own pipeline instead — the "which PR agency should we hire" question, and the entity signals that decide whether an answer engine can describe your firm correctly when a buyer asks it for options. Most agencies start client-side because it is billable to the client, then bring their own site in once they can see what the measurement actually shows.
It measures a different surface. Traditional PR SEO reports links and rankings, and both are now weak proxies: Google's own spam policies list optimised anchor text in press releases as link spam, so wire links do not build authority — that is Google's position, not ours. The published generative-engine research (Aggarwal et al., KDD 2024, a ten-thousand-query benchmark) found the tactics that actually move AI visibility are adding statistics, citing sources, and adding quotations, and that lower-authority domains gain the most from them. AEO work builds that structure and then measures citation directly, rather than inferring it from rankings.
No. A release optimised in isolation has nothing measuring whether it was cited and nowhere to send the authority it earns, which produces exactly the unattributed pickup that made the category look broken in the first place. We do the format work, but inside a programme that also carries the entity layer and the citation measurement. If the format work is genuinely all you need, our guide to formatting a release for AI citation gives the method away for free.
We do not publish a standard figure, because the starting citation position, the newsroom setup, the release cadence, and the strength of the underlying news all change the outcome. Before work starts we agree what counts as a citation, record the current baseline across a fixed prompt set and analytics, and set reporting milestones from there.
Book a discovery call. We will map your market, channels, and first 90-day priorities.